Multiple ergodic averages for commuting multiplicative actions

arXiv:2609.00405 · math.DS · Submitted 2026-08-31 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.

Ada: I'm Ada.

Emmy: And I'm Emmy, here with Kolya.

Ada: We're looking at the paper: "Multiple ergodic averages for commuting multiplicative actions".

Emmy: Here's the gist of the paper. The paper studies the convergence of multiple ergodic averages involving several commuting multiplicative actions and proves that for finitely generated systems, these averages converge in norm.

Ada: Multiple Ergodic Averages in the Multiplicative Setting The work investigates the convergence of multiple ergodic averages involving several commuting multiplicative actions, proving that for finitely generated systems, these averages converge in norm.

Emmy: The authors study the convergence of averages like 1/N X N n=1 S1,nF1 ·...

Ada: · Sl,nFl and prove it settles a conjecture of Frantzikinakis.

Title and authors: Ada: So we're looking at this paper now, "Multiple ergodic averages for commuting multiplicative actions". It looks like they're tackling how several different multiplicative actions behave when you take multiple ergodic averages.

Emmy: Yeah, it’s about those specific sequences of functions and transformations where the actions commute. They are proving that these averages settle a conjecture by Frantzikinakis.

Ada: Exactly, and the big deal here is they prove this convergence in norm for finitely generated systems. That’s a strong statement because norm convergence is much stronger than just almost sure convergence, which we see in some other areas.

Kolya: For me, what's interesting is the machinery they use to get that control over the averages, because typically proving these kinds of bounds involves very delicate tail estimates.

Ada: Right. They introduce a new way to control those averages by generalizing Kátai’s orthogonality criterion, which gives them this box seminorm control. And then they bring in magic extensions from Host’s work to help simplify things.

Emmy: That whole setup is really clever because it allows them to use an induction on complexity of the initial averages, which ultimately reduces the problem down to those well-known mean convergence results.

Ada: So, if I'm thinking about what this means for us, it suggests a systematic way to handle these complicated recurrence problems in multiplicative settings without needing a completely new proof from scratch every time.

Kolya: What I see is that the complexity reduction procedure they develop is quite algorithmic, and it systematically eliminates actions with identity transformations in certain positions until the system is reduced to something simpler.

Emmy: That iterative process, using that complexity vector c equals c1, c2, c3 to quantify the action structure, seems like a very structured way to manage the problem.

Ada: Right. And they show that this reduction can be done by repeatedly applying Theorem four point eight and passing to extensions until you get down to a system where convergence is already known by Theorem A.

Kolya: I wonder about the limitations here, because they mention that the complexity vector itself helps quantify the structure, but it doesn't tell us much about whether this reduction process is always guaranteed to terminate in a finite number of steps for every system.

Emmy: That’s a valid point. The paper establishes that any complexity vector can be reduced to one where convergence is known, which implies the procedure works, but the paper doesn't explicitly state if there are any pathological cases where it might loop or take an infinite amount of time.

Ada: Right. So, to sum up the main improvement they propose is this combination: using the box seminorm control from Kátai's criterion with magic extensions to systematically reduce the problem based on a complexity vector.

Kolya: And that reduction leads to bounding the average by a box seminorm of the initial function, which turns out to be equivalent to a power of that seminorm, which is what guarantees convergence when you have measurability properties.

Emmy: It’s neat how they connect this back to Theorem A for multiplicative systems, showing that any complexity vector can eventually be reduced to one where the average converges in the L2(mu) space.

Ada: So, we've seen how they tackle these multiple ergodic averages involving commuting multiplicative actions and what tools they use to get norm convergence. Now let's look at what they suggest for next steps and open questions.

Emmy: They definitely leave some interesting avenues open, like conjecture one point three about multiplicative actions that might not commute but generate a nilpotent group, which is a big structural question for them <ref:2609.00405#pg2>.

Kolya: And I think the question they ask about joint ergodicity involving polynomial sequences, specifically when the averages of the form En in N F T omega times G in N T2 omega converges for every function F and G, is important because it connects multiplicative actions with additive ones.

Ada: That interaction between additive and multiplicative systems is a key area they are exploring further, and they introduce joint pretentious ergodicity to study those mixed averages.

Emmy: It seems like the next step for this kind of work is to really explore these interactions between the two types of systems, especially how polynomial sequences behave under these combined actions.

Ada: So, to wrap up on this paper, they successfully proved norm convergence for multiple ergodic averages of commuting finitely generated multiplicative systems using a complexity reduction strategy based on Kátai's criterion and magic extensions.

Kolya: And the main caveat is that while the procedure works for reducing complexity, we still need more information about whether this reduction always terminates in a finite number of steps across all possible initial systems.

Emmy: That’s right. The paper shows a powerful framework, but those open questions about non-commuting actions and joint ergodicity show where the next line of research needs to go.

Ada: So, that's our take on "Multiple ergodic averages for commuting multiplicative actions" and how it sets up the next set of problems in this area.

The paper's summary: Ada: So, to recap, this paper is about proving that when you have several different multiplicative actions that all commute—meaning they don't interfere with each other—the multiple ergodic averages still settle down and converge in norm.

Emmy: Yeah, it’s about those sequences of functions and transformations where the actions commute. They are proving that these averages settle a conjecture by Frantzikinakis.

Ada: Exactly, and the big deal here is they prove this convergence in norm for finitely generated systems. That’s a strong statement because norm convergence is much stronger than just almost sure convergence, which we see in some other areas.

Kolya: For me, what's interesting is the machinery they use to get that control over the averages, because typically proving these kinds of bounds involves very delicate tail estimates.

Ada: Right. They introduce a new way to control those averages by generalizing Kátai’s orthogonality criterion, which gives them this box seminorm control. And then they bring in magic extensions from Host’s work to help simplify things.

Emmy: That whole setup is really clever because it allows them to use an induction on complexity of the initial averages, which ultimately reduces the problem down to those well-known mean convergence results.

Ada: So, if I'm thinking about what this means for us, it suggests a systematic way to handle these complicated recurrence problems in multiplicative settings without needing a completely new proof from scratch every time.

Kolya: What I see is that the complexity reduction procedure they develop is quite algorithmic, and it systematically eliminates actions with identity transformations in certain positions until the system is reduced to something simpler.

Emmy: That iterative process, using that complexity vector c equals c1, c2, c3 to quantify the action structure, seems like a very structured way to manage the problem.

Ada: Right. And they show that this reduction can be done by repeatedly applying Theorem four point eight and passing to extensions until you get down to a system where convergence is already known by Theorem A.

Kolya: I wonder about the limitations here, because they mention that the complexity vector itself helps quantify the structure, but it doesn't tell us much about whether this reduction process is always guaranteed to terminate in a finite number of steps for every system.

Emmy: That’s a valid point. The paper establishes that any complexity vector can be reduced to one where convergence is known, which implies the procedure works, but the paper doesn't explicitly state if there are any pathological cases where it might loop or take an infinite amount of time.

Ada: Right. So, to sum up the main improvement they propose is this combination: using the box seminorm control from Kátai's criterion with magic extensions to systematically reduce the problem based on a complexity vector.

Kolya: And that reduction leads to bounding the average by a box seminorm of the initial function, which turns out to be equivalent to a power of that seminorm, which is what guarantees convergence when you have measurability properties.

Emmy: It’s neat how they connect this back to Theorem A for multiplicative systems, showing that any complexity vector can eventually be reduced to one where the average converges in the L2 mu space.

Ada: So, we've seen how they tackle these multiple ergodic averages involving commuting multiplicative actions and what tools they use to get norm convergence. Now let's look at what they suggest for next steps and open questions.

Emmy: They definitely leave some interesting avenues open, like conjecture one point three about multiplicative actions that might not commute but generate a nilpotent group, which is a big structural question for them.

Kolya: And I think the question they ask about joint ergodicity involving polynomial sequences, specifically when the averages of the form En in N F T omega times G in N T2 omega converges for every function F and G, is important because it connects multiplicative actions with additive ones.

Ada: That interaction between additive and multiplicative systems is a key area they are exploring further, and they introduce joint pretentious ergodicity to study those mixed averages.

Emmy: It seems like the next step for this kind of work is to really explore these interactions between the two types of systems, especially how polynomial sequences behave under these combined actions.

Ada: So, that's our take on this paper and how it sets up the next set of problems in this area.

The paper's improvements: Ada: So, to recap, they didn't just prove convergence; they showed how to systematically reduce any complex collection of commuting multiplicative actions down to a simpler one where we already know the answer.

Emmy: Yeah, it’s about that complexity vector c equals c1, c2, c3. It’s like a structural map for the set of transformations S.

Ada: Exactly. They use that vector to guide an iterative construction that eliminates actions with identity transformations in specific positions until all actions are eliminated or reduced to a known convergent case like the average of the form En in N SnF which converges by Theorem A.

Kolya: I mean, from a probabilistic standpoint, it’s interesting because we usually worry about tail probabilities, but this procedure gives us control over those tails by linking them to box seminorms.

Emmy: That’s right. The gains come from bounding the original averages by box seminorms using a variant of Kátai's orthogonality criterion, and then refining those bounds using magic extensions to reduce the problem to simpler averages through an induction on complexity.

Ada: It means we can take a really complicated set of commuting actions and turn it into something manageable just by following that reduction algorithm, which is very useful when you’re dealing with systems that are hard to analyze directly.

Kolya: I wonder if there are any cases where this reduction takes an infinite number of steps, because if it does, the whole method doesn't give us a definitive answer for those systems.

Emmy: That’s the caveat. The paper establishes that any complexity vector can be reduced to one where convergence is known, which implies the procedure works for reducing complexity, but they don't explicitly state if there are any pathological cases where it might loop or take an infinite amount of time.

Ada: So, the improvement here is really the robustness of the method; you get a systematic framework that takes any system and forces it down to a known convergent state through this structured reduction process.

Kolya: And I think what this means for us is that we can tackle recurrence problems in multiplicative systems by using tools from complexity theory and measure theory in a way that’s much more organized than just looking at the raw averages.

Emmy: Exactly. It gives us an algebraic structure to understand how these different multiplicative actions interact, which was something the Host-Kra factors haven't fully adapted to in the general case before.

Ada: So, we’ve seen how they tackle multiple ergodic averages involving commuting multiplicative actions and what tools they use to get norm convergence. Now let's look at what they suggest for next steps and open questions.

Conclusion: Tom: So, to wrap up, this paper proves that we can use these complexity reduction tools to get norm convergence for multiple ergodic averages involving commuting multiplicative actions of finitely generated systems.

Ada: That's right. It’s a big step because it shows we have a reliable way to handle the interplay between additive and multiplicative dynamics when they are well-behaved and commute.

Emmy: Yeah, it settles that conjecture by Frantzikinakis by showing the structure behind these averages is constrained enough for this kind of algorithmic reduction to work.

Kolya: From my side, the main thing is that this method gives us a way to connect recurrence problems in multiplicative systems back to those cleaner L2 convergence results we already have established for simpler cases.

Ada: Exactly. It’s not just about getting a yes or no on convergence; it’s about giving us a systematic procedure that works across different types of commuting actions.

Emmy: The implication is that we can now use the complexity vector to categorize these systems and predict how complex they are before we even start computing anything.

Kolya: I just hope this complexity reduction doesn't have any hidden traps where the number of steps required grows too fast for practical application on larger systems.

Ada: That’s a fair concern. The paper does flag that the procedure reduces complexity, but it doesn't guarantee a finite number of steps for every possible initial system we can imagine.

Emmy: It’s an open question whether this termination is always guaranteed, which brings us right back to those other hard questions about non-commuting actions and joint ergodicity.

Kolya: Yeah, that’s the next big thing for researchers—finding out if these ideas extend when the actions don't commute or when you mix them with polynomial sequences.

Ada: Indeed. The paper on multiple ergodic averages for commuting multiplicative actions gives us a solid foundation before we tackle those more challenging interactions between different types of dynamics.

Dimitrios Charamaras, Andreas Koutsogiannis, Konstantinos Tsinas

Swiss National Science Foundation · CRM-ISM Postdoctoral Fellowship

math.DS

Submitted: 2026-08-31

Updated: 2026-11-06

Comments: 39 pages

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 91/100

The gist: The paper studies the convergence of multiple ergodic averages involving several commuting multiplicative actions and proves that for finitely generated systems, these averages converge in norm.

Key concepts

Multiple Ergodic Averages
These are mathematical averages applied across multiple functions or transformations simultaneously. The paper studies how these averages behave when you have several commuting multiplicative actions, which is a specific type of transformation where the actions interact in a predictable way.
Complexity Vector (c1, c2, c3)
This vector is a tool used to measure the structural complexity of the collection of transformations involved. It quantifies aspects like how many positions exist, the minimum number of non-identity transformations at any spot, and how often those transformations appear together.
Magic Extensions
These are mathematical tools used to simplify problems by ensuring that certain factors derived from the system have a very simple description. By applying magic extensions, the authors can reduce the complexity of a problem involving many actions to one involving fewer or simpler actions.
Box Seminorm Control
This is a method used to bound the size or magnitude of an ergodic average. The paper introduces a new generalization of Kátai's criterion to achieve this control, which is crucial for proving that the averages converge in norm.

Terminology

Summary

The paper studies the convergence of multiple ergodic averages involving several commuting multiplicative actions and proves that for finitely generated systems, these averages converge in norm. p. 0

Where the paper says it

We study the convergence of multiple ergodic averages involving several commuting multiplicative actions. We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · .

Multiple Ergodic Averages in the Multiplicative Setting

The work investigates the convergence of multiple ergodic averages involving several commuting multiplicative actions, proving that for finitely generated systems, these averages converge in norm. The authors study the convergence of averages like 1/N X N n=1 S1,nF1 ·. p. 0

Where the paper says it

We study the convergence of multiple ergodic averages involving several commuting multiplicative actions. We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · .

.. · Sl,nFl and prove it settles a conjecture of Frantzikinakis. p. 0
Where the paper says it

We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · . · Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis.

p. 0
Where the paper says it

We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · . · Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis.

p. 0
Where the paper says it

We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · . · Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis.

p. 0
Where the paper says it

We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · . · Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis.

p. 0
Where the paper says it

We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · . · Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis.

They introduce a novel generalization of Kátai’s orthogonality criterion to obtain box seminorm control, utilizing the machinery of magic extensions originating from Host's work.

Simplification via Additive Systems and Box Seminorms

The proof strategy involves translating the convergence problems into ones involving additive systems with completely additive functions in the iterates, which allows access to box seminorms that can bound the original averages. The authors show that it suffices to establish convergence in a special setting by reducing the problem to an average of lower complexity, using a notion of complexity vector (c1, c2, c3). p. 5

Where the paper says it

Naturally, we say that an average with complexity vector (c1, c2, c3) has lower complexity than an average with complexity vector (c ′ 1 , c′ 2 , c′ 3 ), when (c1, c2, c3) ≺ (c ′ 1 , c′ 2 , c′ 3 ). In the previous example, we started with an average with vector (3, 2, 1) and reduced our problem to an average with complexity vector (3, 1, 2).

This reduction procedure involves applying Theorem 4. p. 5
Where the paper says it

The actual definition of (c1, c2, c3) is slightly more complex for technical reasons, but it conveys the general structure of our argument. We prove that our procedure seminorm control → pass to magic extension → construct new average can be iterated until we reach an average of complexity corresponding to Theorem A.

8 and repeatedly passing to extensions until the collection of actions is reduced to one for which convergence is known. p. 0
Where the paper says it

· Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis. Our methods rely on a novel generalization of Kátai’s orthogonality criterion that allows us to obtain box seminorm control, the machinery of magic extensions originating in the work of Host, and a delicate induction on the complexity of the initial averages that ultimately reduces our problem to well-known mean convergence results.

The Complexity Reduction Procedure

The complexity reduction procedure is an algorithmic method that reduces the problem by iteratively replacing the original average with a simpler one, decreasing its complexity vector. This procedure involves two steps at each stage: first, obtaining a seminorm control through orthogonality criterion applications, and second, reducing the problem to convergence in a new collection of actions via magic extensions. p. 0

Where the paper says it

· Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis. Our methods rely on a novel generalization of Kátai’s orthogonality criterion that allows us to obtain box seminorm control, the machinery of magic extensions originating in the work of Host, and a delicate induction on the complexity of the initial averages that ultimately reduces our problem to well-known mean convergence results.

The process continues until the complexity vector reaches (1, 1, 1), which corresponds to an average whose convergence is known by Theorem A. p. 5
Where the paper says it

The actual definition of (c1, c2, c3) is slightly more complex for technical reasons, but it conveys the general structure of our argument. We prove that our procedure seminorm control → pass to magic extension → construct new average can be iterated until we reach an average of complexity corresponding to Theorem A.

Final Convergence and Joint Ergodicity

The final step involves bounding the resulting average by a box seminorm, which is then related to the box seminorm of the initial function with respect to a collection of transformations. This bound is obtained by applying Theorem 4. p. 4

Where the paper says it

That is, we want to prove convergence for 1 N X N n=1 T a1(n) 1 T a2(n) 2 T a3(n) 3 F1 · R a1(n) 1 R a2(n) 2 R a3(n) 3 F2 for some commuting measure-preserving maps T1, T2, T3, R1, R2, R3. Applying our orthogonality criterion, we can bound this average by the box seminorm3 |||F1|||T1,T1R −1 1 .

8 (3) and utilizing the structure provided by magic extensions. p. 4
Where the paper says it

However, the factor corresponding to this seminorm does not have any beneficial structure in general. Nonetheless, Host observed in [26] that any system always has an extension (called a “magic extension”), in which the corresponding factor admits a helpful description.

The overall result establishes that convergence for a collection of actions implies convergence for another collection with lower complexity, leading to the final conclusion that any complexity vector can be eventually reduced to (1, 1, 1). p. 5
Where the paper says it

Naturally, we say that an average with complexity vector (c1, c2, c3) has lower complexity than an average with complexity vector (c ′ 1 , c′ 2 , c′ 3 ), when (c1, c2, c3) ≺ (c ′ 1 , c′ 2 , c′ 3 ). In the previous example, we started with an average with vector (3, 2, 1) and reduced our problem to an average with complexity vector (3, 1, 2).

Open Problems and Generalizations

The paper concludes by posing several open problems, including Conjecture 1.3 regarding multiplicative actions that do not necessarily commute generating a nilpotent group. p. 5

Where the paper says it

Let (X, X , µ) be a probability space, and S1, . , Sℓ be finitely generated multiplicative actions generating a nilpotent group.

p. 5
Where the paper says it

Let (X, X , µ) be a probability space, and S1, . , Sℓ be finitely generated multiplicative actions generating a nilpotent group.

Another question asks under what assumptions the averages of the form En∈[N]F(Tomega(n)x) · G(T2omega(n)x) converge for every F, G in C(X) and every x ∈ X. p. 6
Where the paper says it

Under what assumptions do we have that the averages 1 N X N n=1 F(T Ω(n)x) · G(T 2Ω(n)x) converge for every F, G ∈ C(X) and every x ∈ X? Another interesting question arises from the interaction between additive and multiplicative systems.

p. 6
Where the paper says it

Under what assumptions do we have that the averages 1 N X N n=1 F(T Ω(n)x) · G(T 2Ω(n)x) converge for every F, G ∈ C(X) and every x ∈ X? Another interesting question arises from the interaction between additive and multiplicative systems.

Furthermore, Conjecture 1. p. 0
Where the paper says it

We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · . · Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis.

5 explores the interaction between additive and multiplicative systems regarding the convergence of mixed averages involving polynomial sequences. p. 6
Where the paper says it

Under what assumptions do we have that the averages 1 N X N n=1 F(T Ω(n)x) · G(T 2Ω(n)x) converge for every F, G ∈ C(X) and every x ∈ X? Another interesting question arises from the interaction between additive and multiplicative systems.

The paper also introduces the notion of joint pretentious ergodicity to study these interactions. p. 6
Where the paper says it

In simplistic terms, joint ergodicity is the phenomenon of convergence of ergodic averages to the product of the integrals of the functions involved. Since we are using additive averages to study multiplicative actions, the correct notion to use is that of pretentious ergodicity, introduced in [8].

p. 6
Where the paper says it

In simplistic terms, joint ergodicity is the phenomenon of convergence of ergodic averages to the product of the integrals of the functions involved. Since we are using additive averages to study multiplicative actions, the correct notion to use is that of pretentious ergodicity, introduced in [8].

Notation and Definitions

The authors define various notations for sets, functions, and averages, including P-free numbers QP and logarithmic averages over sets A. They specify asymptotic notation such as O(g(x)) and o(x) → ∞(1). p. 7

Where the paper says it

For u ∈ N0 and z ∈ C, we define C u z := ( z, if u is even, z, if u is odd. In addition, we use the standard asymptotic notation.

p. 7
Where the paper says it

For u ∈ N0 and z ∈ C, we define C u z := ( z, if u is even, z, if u is odd. In addition, we use the standard asymptotic notation.

The complexity vector c = (c1, c2, c3) is formally defined based on the number of positions, the minimum number of non-identity transformations in a position, and the minimum multiplicity among those transformations. p. 5
Where the paper says it

5) and m = 3 in both of the previous examples). ii) We identify the index i that has the least number of distinct transformations Ti,j (thus, the most repetitions), and we let c2 be the number of distinct transformations in this position.

p. 5
Where the paper says it

5) and m = 3 in both of the previous examples). ii) We identify the index i that has the least number of distinct transformations Ti,j (thus, the most repetitions), and we let c2 be the number of distinct transformations in this position.

The definition of factors and extensions for measure-preserving systems is also detailed. p. 7
Where the paper says it

Background 2. Measure-preserving systems, factors and extensions.

p. 7
Where the paper says it

Background 2. Measure-preserving systems, factors and extensions.

Background on Multiplicative Actions

The paper discusses multiplicative systems as quadruples (X, X, µ, S) where Snm = Sn ◦ Sm for any m, n ∈ N. Simple examples of multiplicative systems include rotations by multiplicative functions and dilations on the torus. p. 1

Where the paper says it

If we assume that the transformations Sn are invertible, then they naturally induce a (Q+, ×)-action on (X, X , µ) by Sn/m = Sn ◦ S −1 m . Simple examples of multiplicative systems include: i) rotations by multiplicative functions: if f: N → S 1 is a completely multiplicative function, then we define the map Sn : S 1 → S 1 by Snz = f(n)z, which preserves the Lebesgue measure on S 1 .

p. 1
Where the paper says it

If we assume that the transformations Sn are invertible, then they naturally induce a (Q+, ×)-action on (X, X , µ) by Sn/m = Sn ◦ S −1 m . Simple examples of multiplicative systems include: i) rotations by multiplicative functions: if f: N → S 1 is a completely multiplicative function, then we define the map Sn : S 1 → S 1 by Snz = f(n)z, which preserves the Lebesgue measure on S 1 .

The study of recurrence results for multiplicative actions has been sparse, though results in this field are connected to partition regularity results for homogeneous quadratic forms, [17], [16], [20]. p. 2
Where the paper says it

◦ Sp ak k . Multiplicative systems have recently gained prominence due to their connections to partition regularity results for homogeneous quadratic forms [18], [17], [16], [20].

p. 2
Where the paper says it

◦ Sp ak k . Multiplicative systems have recently gained prominence due to their connections to partition regularity results for homogeneous quadratic forms [18], [17], [16], [20].

The structure theory of Host-Kra factors has not been fully adapted to multiplicative actions in the general case. p. 2
Where the paper says it

Using a multiplicative averaging scheme, some results from the additive setting admit analogs in the multiplicative setting. Still, the structure theory of HostKra factors has not been adapted to multiplicative actions (it is restricted in the general case to finitely generated abelian group actions [24] or to lower-order cases [28]) and most results involving multiple averages remain unknown.

p. 2
Where the paper says it

Using a multiplicative averaging scheme, some results from the additive setting admit analogs in the multiplicative setting. Still, the structure theory of HostKra factors has not been adapted to multiplicative actions (it is restricted in the general case to finitely generated abelian group actions [24] or to lower-order cases [28]) and most results involving multiple averages remain unknown.

Key Results on Multiplicative Averages

Theorem A states that for a finitely generated multiplicative system, the averages 1/N X N n=1 SnF converge in L2(µ). The main result, Theorem 1. p. 3

Where the paper says it

Theorem A is a consequence of the Pretentious Mean Ergodic Theorem [8, Theorem A], an analogue of the von Neumann mean ergodic theorem for finitely generated multiplicative systems, which also gives an explicit description of the limit. Our main result is an extension of Theorem A to the setting of multiple ergodic averages.

1, extends this to multiple ergodic averages of commuting finitely generated multiplicative systems. p. 0
Where the paper says it

We study the convergence of multiple ergodic averages involving several commuting multiplicative actions. We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · .

The authors prove that for a commuting finitely generated multiplicative system (X, X, µ, S1,. p. 0
Where the paper says it

We study the convergence of multiple ergodic averages involving several commuting multiplicative actions. We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · .

.., Sl), the averages 1/N X N n=1 S1,nF1 ·. p. 0
Where the paper says it

We study the convergence of multiple ergodic averages involving several commuting multiplicative actions. We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · .

.. · Sl,nFl converge in L2(µ). p. 0
Where the paper says it

We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · . · Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis.

Convergence of Specific Averages

The paper demonstrates convergence for specific complex averages by applying the orthogonality criterion to bound them by box seminorms. For example, the average 1/N X N n=1 T a1(n) 1 T a2(n) 2 F · R a1(n) 1 R a2(n) 2 G · R a1(n) 1 T a2(n) 2 H · R a2(n) 2 W converges in L 2(µ), provided the system is magic with respect to T1, T1R−1 and has certain properties. p. 4

Where the paper says it

Applying our orthogonality criterion, we can bound this average by the box seminorm3 |||F1|||T1,T1R −1 1 . In later sections, we provide an example that illustrates how we arrive at these types of bounds.

The Role of Magic Extensions

Magic extensions are used to extract information from seminorm bounds by ensuring that the corresponding factor admits a simple description. The authors show that for any system, an extension exists where certain factors admit a helpful description, which reduces the complexity of the problem. p. 1

Where the paper says it

Later, Austin [1] gave a genuinely ergodic theoretic proof, followed by Host [26], who gave another proof inspired by Austin’s arguments. A common idea in these last two articles was the construction of an extension of the original system, wherein certain factors that control the averages admit a simple description that reduces the complexity of the problem.

This allows them to reduce the original convergence problem to one involving simpler averages through induction on complexity. p. 0
Where the paper says it

· Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis. Our methods rely on a novel generalization of Kátai’s orthogonality criterion that allows us to obtain box seminorm control, the machinery of magic extensions originating in the work of Host, and a delicate induction on the complexity of the initial averages that ultimately reduces our problem to well-known mean convergence results.

The Complexity Vector and Reduction

The complexity vector c = (c1, c2, c3) is used to quantify the structure of the collection of actions S. The reduction procedure systematically reduces this vector by eliminating actions that have the identity transformation in a certain position. p. 5

Where the paper says it

In the average after the reduction, we note that c2 has reduced to 1. Indeed, there is only one transformation corresponding to the position a1 (namely, Re1).

This iterative process ensures that convergence for a collection with complexity vector c can be deduced from convergence for another collection with complexity vector c' < c via finitely many applications of the reduction procedure.

The Final Bound

The final step in the proof involves bounding the average by a box seminorm derived from the transformations appearing in the product of actions. This bound is shown to be equivalent to a power of that specific box seminorm, which implies convergence when F is measurable with respect to the invariant factor. p. 4

Where the paper says it

If G1 is orthogonal to the factor corresponding to the seminorm ||| · |||Te1,Te1Re−1 1 , then the average converges to zero by our seminorm bound, and we are done. Otherwise, we may assume that our function G1 is measurable with respect to this factor.

The entire procedure concludes by showing that any complexity vector can be reduced to (1, 1, 1), which corresponds to an average whose convergence is known by Theorem A. p. 5
Where the paper says it

Naturally, we say that an average with complexity vector (c1, c2, c3) has lower complexity than an average with complexity vector (c ′ 1 , c′ 2 , c′ 3 ), when (c1, c2, c3) ≺ (c ′ 1 , c′ 2 , c′ 3 ). In the previous example, we started with an average with vector (3, 2, 1) and reduced our problem to an average with complexity vector (3, 1, 2).

Summary of Convergence for Multiple Averages

The paper successfully establishes the convergence of multiple ergodic averages involving commuting finitely generated multiplicative actions in norm. The core technical achievement lies in the novel generalization of Kátai’s criterion to obtain box seminorm control, coupled with a reduction procedure based on magic extensions and complexity vectors. p. 0

Where the paper says it

· Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis. Our methods rely on a novel generalization of Kátai’s orthogonality criterion that allows us to obtain box seminorm control, the machinery of magic extensions originating in the work of Host, and a delicate induction on the complexity of the initial averages that ultimately reduces our problem to well-known mean convergence results.

This methodology provides a robust framework for tackling complex recurrence problems in this setting. p. 2
Where the paper says it

They arise naturally in combinatorial problems that involve sets that are multiplicatively large. Recurrence results for multiplicative actions have been studied [12] (see also [9] and [32]), though results in this field have been sparse.

p. 2
Where the paper says it

They arise naturally in combinatorial problems that involve sets that are multiplicatively large. Recurrence results for multiplicative actions have been studied [12] (see also [9] and [32]), though results in this field have been sparse.

References

[1] T. Austin. On the norm convergence of non-conventional ergodic averages

[2] D. Berend and V. p. 3

Where the paper says it

While the theorem might be true if we assume that the actions generate a nilpotent group, it does not hold for arbitrary collections of actions. To see this, one can modify a construction of Berend [2].

p. 3
Where the paper says it

While the theorem might be true if we assume that the actions generate a nilpotent group, it does not hold for arbitrary collections of actions. To see this, one can modify a construction of Berend [2].

Bergelson. Joint ergodicity and mixing

[3] D. Berend and V. Bergelson. Jointly ergodic measure-preserving transformations

[4] V. Bergelson, Weakly mixing PET

[5] V. Bergelson and F. p. 5

Where the paper says it

Nevertheless, we claim that the complexity of the average decreased. This is reminiscent of the PET induction scheme in the additive setting (via the van der Corput inequality), originally introduced by Bergelson in [4].

K. Richter. Dynamical generalizations of the prime number theorem and disjointness of additive and multiplicative semigroup actions

[6] J. Bourgain, Double recurrence and almost sure convergence

[7] J. Bourgain, P. p. 1

Where the paper says it

Simple examples of multiplicative systems include: i) rotations by multiplicative functions: if f: N → S 1 is a completely multiplicative function, then we define the map Sn : S 1 → S 1 by Snz = f(n)z, which preserves the Lebesgue measure on S 1 . 1The pointwise convergence of these averages is a notorious open problem, only known in the case ℓ = 2 by work of Bourgain [6].

Sarnak, and T. Ziegler. Disjointness of Möbius from horocycle flows

[8] D. Charamaras. p. 0

Where the paper says it

MULTIPLE ERGODIC AVERAGES FOR COMMUTING MULTIPLICATIVE ACTIONS DIMITRIOS CHARAMARAS, ANDREAS KOUTSOGIANNIS, AND KONSTANTINOS TSINAS Abstract. We study the convergence of multiple ergodic averages involving several commuting multiplicative actions.

Mean value theorems in multiplicative systems and joint ergodicity of additive and multiplicative actions

[9] D. Charamaras, A. p. 0

Where the paper says it

MULTIPLE ERGODIC AVERAGES FOR COMMUTING MULTIPLICATIVE ACTIONS DIMITRIOS CHARAMARAS, ANDREAS KOUTSOGIANNIS, AND KONSTANTINOS TSINAS Abstract. We study the convergence of multiple ergodic averages involving several commuting multiplicative actions.

Mountakis, and K. Tsinas. On multiplicative recurrence along linear patterns

[10] J.-P. Conze and E. p. 1

Where the paper says it

converge in L 2 (µ), at least for ℓ ≥ 3. 1 The mean convergence of the averages was a major open problem until 2005, when it was resolved by Host and Kra [27] (building on work of Conze and Lesigne [10], who verified the conjecture for ℓ = 3 and totally ergodic systems) with a different proof by Ziegler [41].

Lesigne. Théorèmes ergodiques pour des mesures diagonales

[11] S. Donoso, A. Koutsogiannis, B. p. 0

Where the paper says it

MULTIPLE ERGODIC AVERAGES FOR COMMUTING MULTIPLICATIVE ACTIONS DIMITRIOS CHARAMARAS, ANDREAS KOUTSOGIANNIS, AND KONSTANTINOS TSINAS Abstract. We study the convergence of multiple ergodic averages involving several commuting multiplicative actions.

Kuca, W. Sun, and K. Tsinas. Seminorm estimates and joint ergodicity for pairwise independent Hardy sequences

[12] S. Donoso, A. N. Le, J. Moreira, and W. Sun. Additive averages of multiplicative correlation sequences and applications

[13] N. Frantzikinakis. Decomposition results for multiplicative actions and applications

[14] N. Frantzikinakis. Multiple recurrence and convergence for Hardy sequences of polynomial growth

[15] N. Frantzikinakis. Some open problems on multiple ergodic averages

[16] N. Frantzikinakis. Partition regularity of homogeneous quadratics: Current trends and challenges

[17] N. Frantzikinakis, O. p. 0

Where the paper says it

We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · . · Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis.

p. 0
Where the paper says it

We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · . · Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis.

Klurman, and J. Moreira. Partition regularity of generalized pythagorean pairs

[18] N. Frantzikinakis, O. Klurman, and J. Moreira. Partition regularity of Pythagorean pairs

[19] N. Frantzikinakis and B. p. 0

Where the paper says it

We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · . · Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis.

Kuca. Ergodic averages for sparse corners

[20] N. Frantzikinakis and A. p. 0

Where the paper says it

We prove that for finitely generated systems, the averages 1 N XN n=1 S1,nF1 · . · Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis.

Mountakis. Recurrence for pretentious systems along generalized Pythagorean triples

[21] H. Furstenberg. p. 0

Where the paper says it

Introduction 1. The study of the limiting behavior of multiple ergodic averages has been a major field of study originating from Furstenberg’s influential work on Szemerédi’s theorem.

p. 0
Where the paper says it

Introduction 1. The study of the limiting behavior of multiple ergodic averages has been a major field of study originating from Furstenberg’s influential work on Szemerédi’s theorem.

Ergodic behavior of diagonal measures and a theorem of Szemerédi on arithmetic progressions

[22] H. Furstenberg. Recurrence in Ergodic Theory and Combinatorial Number Theory

[30] B. Kuca. Joint ergodicity - 40 years on

[31] A. Leibman. Convergence of multiple ergodic averages along polynomials of several variables

[32] S.-K. Leung and C. Táfula. Multiplicative recurrence of Möbius transformations

[34] F. K. Richter. A new elementary proof of the prime number theorem

[35] T. Tao. Norm convergence of multiple ergodic averages for commuting transformations

[36] T. Tao and T. Ziegler. Concatenation theorems for anti-Gowers uniform functions and Host-Kra characteristic factors

[37] H. Towsner. Convergence of diagonal ergodic averages

[40] R. Xiao. Some ergodic theorems involving omega function and their applications

--- Page 1 ---

The study of the limiting behavior of multiple ergodic averages has been a major field of study originating from Furstenberg’s influential work on Szemerédi’s theorem. The mean convergence of the averages was a major open problem until 2005, when it was resolved by Host and Kra [27] (building on work of Conze and Lesigne [10], who verified the conjecture for l = 3 and totally ergodic systems) with a different proof by Ziegler [41]. p. 1

Where the paper says it

converge in L 2 (µ), at least for ℓ ≥ 3. 1 The mean convergence of the averages was a major open problem until 2005, when it was resolved by Host and Kra [27] (building on work of Conze and Lesigne [10], who verified the conjecture for ℓ = 3 and totally ergodic systems) with a different proof by Ziegler [41].

The highlight of the newly introduced tools was not only the proof of the convergence, but also the introduction of some factors that control the behavior of these averages (1. p. 1
Where the paper says it

1 The mean convergence of the averages was a major open problem until 2005, when it was resolved by Host and Kra [27] (building on work of Conze and Lesigne [10], who verified the conjecture for ℓ = 3 and totally ergodic systems) with a different proof by Ziegler [41]. The highlight of the newly introduced tools was not only the proof of the convergence, but the introduction of some factors of the system (X, µ, T) that control the behavior of the averages (1.

2) which have an explicit algebraic structure. p. 1
Where the paper says it

The highlight of the newly introduced tools was not only the proof of the convergence, but the introduction of some factors of the system (X, µ, T) that control the behavior of the averages (1. 2) and which have an explicit algebraic structure.

How Training Signal is Built

The authors construct a complexity vector c = (c1, c2, c3) to quantify the structure of the collection of actions S. This vector is defined by counting the number of positions, the minimum number of non-identity transformations in a position, and the minimum multiplicity among those transformations. p. 5

Where the paper says it

5) and m = 3 in both of the previous examples). ii) We identify the index i that has the least number of distinct transformations Ti,j (thus, the most repetitions), and we let c2 be the number of distinct transformations in this position.

The reduction procedure uses this vector to guide an iterative construction that eliminates actions with identity transformations in specific positions. p. 5
Where the paper says it

For instance, we have c3 = 1 in the first example, since in position a1, we have that both T1, R1 appear once (the same holds for the other positions a2, a3). Similarly, we have that the new average after the reduction has c3 = 2, since in position a1 (where c2 is attained), we have only one transformation Re1, which appears twice.

This process continues until all actions are eliminated or reduced to a known convergent case, such as the average of the form En∈[N]SnF which converges by Theorem A. p. 5
Where the paper says it

Using the relations Te1H1 = H1 and Te1H2 = Re1H2, we have to prove convergence for the averages 1 N X N n=1 Tea2(n) 2 Tea3(n) 3 H1 · Rea1(n) 1 Tea2(n) 2 Tea3(n) 3 H2 · Rea1(n) 1 Rea2(n) 2 Rea3(n) 3 G2. We claim by induction that an average of this form converges for all systems and any choice of functions H1, H2, H3, and then we are done.

Where the Gains Come From

The gains in proving convergence come from bounding the original averages by box seminorms using a variant of Kátai’s orthogonality criterion. These bounds are then refined using magic extensions to reduce the problem to simpler averages, which is achieved through an inductive argument on complexity. p. 0

Where the paper says it

· Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis. Our methods rely on a novel generalization of Kátai’s orthogonality criterion that allows us to obtain box seminorm control, the machinery of magic extensions originating in the work of Host, and a delicate induction on the complexity of the initial averages that ultimately reduces our problem to well-known mean convergence results.

This combination allows for a systematic reduction of the problem until it reaches a known convergent state, effectively proving Theorem 4. p. 0
Where the paper says it

· Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis. Our methods rely on a novel generalization of Kátai’s orthogonality criterion that allows us to obtain box seminorm control, the machinery of magic extensions originating in the work of Host, and a delicate induction on the complexity of the initial averages that ultimately reduces our problem to well-known mean convergence results.

4.

The paper concludes by posing several open problems, including Conjecture 1.3 regarding multiplicative actions that do not necessarily commute generating a nilpotent group. Another question asks under what assumptions the averages of the form En∈[N]F(Tomega(n)x) · G(T2omega(n)x) converge for every F, G in C(X) and every x ∈ X. The paper also introduces the notion of joint pretentious ergodicity to study these interactions. These open questions highlight the ongoing challenges in extending these convergence results to more complex settings. p. 1

Where the paper says it

1 The mean convergence of the averages was a major open problem until 2005, when it was resolved by Host and Kra [27] (building on work of Conze and Lesigne [10], who verified the conjecture for ℓ = 3 and totally ergodic systems) with a different proof by Ziegler [41]. The highlight of the newly introduced tools was not only the proof of the convergence, but the introduction of some factors of the system (X, µ, T) that control the behavior of the averages (1.

The authors define various notations for sets, functions, and averages, including P-free numbers QP and logarithmic averages over sets A. They specify asymptotic notation such as O(g(x)) and o(x) → ∞(1). The complexity vector c = (c1, c2, c3) is formally defined based on the number of positions, the minimum number of non-identity transformations in a position, and the minimum multiplicity among those transformations. The definition of factors and extensions for measure-preserving systems is also detailed. The study of multiplicative actions is connected to partition regularity results for homogeneous quadratic forms. p. 2

Where the paper says it

◦ Sp ak k . Multiplicative systems have recently gained prominence due to their connections to partition regularity results for homogeneous quadratic forms [18], [17], [16], [20].

The paper uses standard notation such as N for natural numbers and P for primes. p. 7
Where the paper says it

Similar conjectures can be formulated for more complicated averages, where each action is evaluated along sequences other than a(n) = n, but we do not state the problem in this generality. We use the standard notation N for the natural numbers, N0 := N∪ {0}, Z for the integers, Q for the rational numbers, Q+ for the positive rationals, P for the primes, R for the real numbers, C for the complex numbers, and S 1 for the complex unit circle.

The paper discusses multiplicative systems as quadruples (X, X, µ, S) where Snm = Sn ◦ Sm for any m, n ∈ N. Simple examples of multiplicative systems include rotations by multiplicative functions and dilations on the torus. The study of recurrence results for multiplicative actions has been sparse, though results in this field are connected to partition regularity results for homogeneous quadratic forms, [17], [16], [20]. The structure theory of Host-Kra factors has not been fully adapted to multiplicative actions in the general case. The paper notes that the natural way to study multiplicative actions is through multiplicative Følner sequences. p. 2

Where the paper says it

Recurrence results for multiplicative actions have been studied [12] (see also [9] and [32]), though results in this field have been sparse. We remark that the natural way to study multiplicative actions is through multiplicative Følner sequences.

The paper successfully establishes the convergence of multiple ergodic averages involving commuting finitely generated multiplicative actions in norm. The core technical achievement lies in the novel generalization of Kátai’s orthogonality criterion to obtain box seminorm control, coupled with a reduction procedure based on magic extensions and complexity vectors. p. 0

Where the paper says it

· Sℓ,nFℓ converge in norm, settling a conjecture of Frantzikinakis. Our methods rely on a novel generalization of Kátai’s orthogonality criterion that allows us to obtain box seminorm control, the machinery of magic extensions originating in the work of Host, and a delicate induction on the complexity of the initial averages that ultimately reduces our problem to well-known mean convergence results.

This methodology provides a robust framework for tackling complex recurrence problems in this setting.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved AI can prove that for finitely generated systems, multiple ergodic averages involving several commuting multiplicative actions converge in norm, settling a conjecture of Frantzikinakis.

  2. Improved AI can utilize a novel generalization of Kátai’s orthogonality criterion to obtain box seminorm control and apply the machinery of magic extensions to reduce the problem to well-known mean convergence results.

  3. Improved AI can perform complexity reduction on multiple ergodic averages by iteratively replacing the average with simpler ones, eventually reducing any collection of actions to a system where convergence is known, such as when all transformations are identity.

  4. Improved AI can solve problems involving joint ergodicity by checking conditions related to whether SiS-1j is pretentiously ergodic for all distinct i and j, and the product S1 x... x Sl is pretentiously ergodic for µ x... x µ [z] times.

  5. Improved AI can analyze the convergence of averages involving polynomial ergodics by combining complexity reduction with the classical PET induction scheme, allowing it to reduce problems to known convergence results via Proposition 4.9.

  6. Improved AI can determine if a collection of actions S is reduced to R(S) after one application of the complexity reduction procedure by examining whether Ti0,1 attains c3 and checking the resulting complexity vector.

Abstract

We study the convergence of multiple ergodic averages involving several commuting multiplicative actions. We prove that for finitely generated systems, the averages 1 over N sum from n=1 to N S 1,nF 1 times times S,nF converge in norm, settling a conjecture of Frantzikinakis. Our proofs feature tools from ergodic theory and probabilistic number theory. We exploit some stability properties of additive functions to establish a novel variant of 's orthogonality criterion, which leads to box seminorm control for our averages. Using the theory of magic extensions introduced by Host and a delicate complexity reduction argument, we eventually reduce our problem to known convergence results.

Sources

Related papers