Averages of sub-additive functions through non-F lner sequences

arXiv:2611.08408 · math.DS, math.GR · Submitted 2026-11-05 · 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: "Averages of sub-additive functions through non-F lner sequences".

Emmy: Here's the gist of the paper. The paper studies averages of sub-additive functions through non-Følner sequences to generalize classical results on sequence entropy for countable amenable groups.

Ada: This work is significant because it removes technical conditions from previous generalizations and proves that every choice of a non-Følner sequence overestimates topological and measure-theoretic entropy by a multiplicative factor related to the "Følnerness" constant C(F).

Emmy: Interpretation of the Følnerness Constant The constant C(F) is defined as sup E⋐G lim sup n→∞ EFn/Fn, and Remark 1.2 states that C(F) equals 1 if and only if F is Følner.

Ada: The paper notes that C(F) can be interpreted as a “Følnerness” constant.

Paper summary: Ada: So to summarize this paper, "Averages of sub-additive functions through non-Følner sequences," the core thesis is that they study averages of sub-additive functions using sequences that are not necessarily Følner.

Emmy: The main claim is that under mild conditions on those functions, they prove that this quantity equals the supremum over those same functions of the limit of the average as you make the set more and more invariant, multiplied by a Følnerness constant C(F).

Kolya: And what makes it matter is this constant C(F), which is one if and only if the sequence is Følner, meaning it measures exactly how far you are from being well-behaved.

Ada: Exactly. The real significance comes from a corollary of this result, which states that every choice of a non-Følner sequence overestimates topological and measure-theoretic entropy by that multiplicative factor C(F) in the positive entropy case.

Emmy: It means if you pick any sequence that isn't Følner, the estimated entropy is inflated by this constant C(F).

Kolya: So for a practical application, what does that tell us about how we estimate complexity in these systems? Is it just an error bound?

Ada: It’s more than just an error bound; it’s a quantified statement about the inherent bias when you use a non-Følner sequence to measure entropy.

Emmy: They're generalizing results from Goodman, Krug, and Newton on the proportionality principle for sequence entropy, and extending that by Gao, Huang, and Zhu to countable amenable groups.

Kolya: So they are taking established ideas about how sequences behave in these settings and putting them into a new framework that handles non-Følner cases more generally.

Conclusion: Ada: Thinking about the title, "Averages of sub-additive functions through non-Følner sequences," it sounds very technical and dense, but the paper’s aim is to make that concept accessible by connecting it to more familiar ideas like sub-additive functions.

Emmy: The authors are tackling a deep topic in dynamical systems—sequence entropy—and they manage to connect it cleanly to the structure of countable amenable groups.

Kolya: The implication for us, as listeners, is that we can use non-Følner sequences and still get meaningful bounds on complexity because this error is explicitly tied to the Følnerness constant C(F).

Ada: So in simple terms, it means if you're working with these groups and you pick a sequence that isn't Følner, you can use the paper’s formula to know exactly how much your estimate will be inflated by C(F).

Emmy: It gives us a precise tool to measure that deviation from the ideal Følner behavior in terms of entropy.

Kolya: It seems like this work provides a clear path forward for generalizing these classical results without getting bogged down in overly specific technical prerequisites that held up earlier attempts.

Universit Paris Cit, Paris, France

math.DS, math.GR

Submitted: 2026-11-05

Updated: 2026-11-05

Comments: 17 pages

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

Importance score: 79/100

The gist: The paper studies averages of sub-additive functions through non-Følner sequences to generalize classical results on sequence entropy for countable amenable groups.

Key concepts

Non-Følner Sequence
This is a specific type of sequence used in the study. A Følner sequence has special properties, but this paper examines sequences that do not possess these properties. They are crucial because they allow the generalization of entropy results beyond classical conditions.
Følnerness Constant C(F)
This constant measures how far a sequence is from being a Følner sequence. If C(F) equals 1, the sequence is Følner. If it's greater than 1, it indicates that the sequence overestimates entropy by that factor.
Sequence Entropy
This concept generalizes classical entropy to sequences of functions or sets. The paper shows how to compute this for topological and measure-theoretic systems, linking them to the group's structure through specific sequence averages.

Terminology

Summary

The paper studies averages of sub-additive functions through non-Følner sequences to generalize classical results on sequence entropy for countable amenable groups. p. 0

Where the paper says it

AVERAGES OF SUB-ADDITIVE FUNCTIONS THROUGH NON-FØLNER SEQUENCES NICANOR CARRASCO-VARGAS Abstract. We study the quantity supϕ∈Φ lim supn→∞ ϕ(Fn)/|Fn|, where F = (Fn)n∈N is a sequence of finite subsets of a countable amenable group, and Φ is a collection of subadditive functions.

This work is significant because it removes technical conditions from previous generalizations and proves that every choice of a non-Følner sequence overestimates topological and measure-theoretic entropy by a multiplicative factor related to the Følnerness constant C(F).

Interpretation of the Følnerness Constant

The constant C(F) is defined as sup E⋐G lim sup n→∞ EFn/Fn, and Remark 1.2 states that C(F) equals 1 if and only if F is Følner. p. 1

Where the paper says it

AVERAGES OF SUB-ADDITIVE FUNCTIONS THROUGH NON-FØLNER SEQUENCES 2 Let G be a countable amenable group and let F(G) be the set of finite nonempty subsets of G. A generalization of Fekete’s Lemma, often called Ornstein-Weiss Lemma, states that when ϕ: F(G) → [0, ∞) is subadditive and translation-invariant then F 7→ 1 |F| ϕ(F) converges to a limit as F becomes more and more invariant.

The paper notes that C(F) can be interpreted as a “Følnerness” constant. p. 0
Where the paper says it

We study the quantity supϕ∈Φ lim supn→∞ ϕ(Fn)/|Fn|, where F = (Fn)n∈N is a sequence of finite subsets of a countable amenable group, and Φ is a collection of subadditive functions. Under mild conditions on Φ we prove that the aforementioned quantity equals the supremum over ϕ ∈ Φ of the limit of ϕ(F)/|F| as F becomes more and more invariant, multiplied by a “Følnerness” constant C(F) ∈ [1, ∞].

Connection to Entropy Invariants

The result implies that every choice of a non-Følner sequence overestimates topological and measure-theoretic entropy by the multiplicative factor C(F) (in the positive entropy case). This is seen as a version of the Ornstein-Weiss Lemma or Fekete’s Lemma for non-Følner sequences. p. 1

Where the paper says it

In particular, every choice of a non-Følner sequence overestimates entropy, where the excess takes the form of a multiplicative factor C(F). Thus this result can be seen as a version of the Ornstein-Weiss Lemma or Fekete’s Lemma for non-Følner sequences, with the caveat that it concerns a family of sub-additive functions instead of individual functions.

Topological and Measure-Theoretic Entropy Generalizations

The paper derives results about sequence entropy by applying Theorem 3.3 to specific families of functions related to topological entropy (using log N(U F)) and measure-theoretic entropy (using Hµ(ξ F)). p. 1

Where the paper says it

Then hϕ = hµ(T, ξ) and h = hµ(T). It is well-known that for the computation of entropy (either topological or measure-theoretic), if we use a non-Følner sequence in the definition, then we may obtain something different from entropy.

It shows that for sequences where either K(A) 0, the sequence entropy equals K(A) · htop(T). p. 1
Where the paper says it

It is well-known that for the computation of entropy (either topological or measure-theoretic), if we use a non-Følner sequence in the definition, then we may obtain something different from entropy. 1 shows that in the positive entropy case, we in fact obtain entropy scaled by the factor C(F) ∈ [1, ∞].

Goodwyn’s Theorem and Boundedness

Theorem 5.1 proves that the measure-theoretic sequence entropy is always bounded by the topological one, sup µ h Aµ(T) ≤ h Atop(T), with equality holding if either K(A) 0. p. 1

Where the paper says it

Then hϕ = hµ(T, ξ) and h = hµ(T). It is well-known that for the computation of entropy (either topological or measure-theoretic), if we use a non-Følner sequence in the definition, then we may obtain something different from entropy.

Proportionality Principle for Sequence Entropy

The paper proves that the proportionality principle holds for sequence entropy, showing that sup ϕ∈Φ lim sup n→∞ ϕ(Fn) s(n) = C(F) · h under certain conditions. This is achieved by analyzing the good and bad parts of the sum in a proof structure. p. 6

Where the paper says it

This should become clear in the rest of the proof. We split the sum in the “good part” and the “bad part” as = lim sup n→∞   X i∈N (n,i) good 1 s(h(n))ϕ(EFh(n) ∩ Tigi) + X i∈N (n,i) bad 1 s(h(n))ϕ(EFh(n) ∩ Tigi)   We will now bound the good part from above.

Topological Sequence Entropy Computation

Theorem 4.5 shows that if either K(A) 0, then h Atop(T) = K(A) · htop(T), where the term at the right is interpreted as zero if K(A) = 0 and htop(T) = ∞. p. 1

Where the paper says it

It is well-known that for the computation of entropy (either topological or measure-theoretic), if we use a non-Følner sequence in the definition, then we may obtain something different from entropy. 1 shows that in the positive entropy case, we in fact obtain entropy scaled by the factor C(F) ∈ [1, ∞].

This is applied to topological sequence entropy using sequences Fn = t1,. p. 0
Where the paper says it

The proportionality principle mentioned in the previous paragraph was first discovered by Krug and Newton [KN] for measure-theoretic sequence entropy, and subsequently generalized by Goodman [G] to topological sequence entropy. A generalization of the same results for actions of countable amenable groups was recently proved by Gao, Huang and Zhu [GHZ], for sequences satisfying a technical condition (see also [E,KK2,KK1,HY]).

.., tn and s(n) = n.

Measure-Theoretic Sequence Entropy Computation

Theorem 4.9 shows that if either K(A) 0, then h Aµ(T) = K(A) · hµ(T), where the term at the right is interpreted as zero if K(A) = 0 and hµ(T) = ∞. p. 1

Where the paper says it

Then hϕ = hµ(T, ξ) and h = hµ(T). It is well-known that for the computation of entropy (either topological or measure-theoretic), if we use a non-Følner sequence in the definition, then we may obtain something different from entropy.

This is applied to measure-theoretic sequence entropy using sequences Fn = t1,. p. 0
Where the paper says it

The proportionality principle mentioned in the previous paragraph was first discovered by Krug and Newton [KN] for measure-theoretic sequence entropy, and subsequently generalized by Goodman [G] to topological sequence entropy. A generalization of the same results for actions of countable amenable groups was recently proved by Gao, Huang and Zhu [GHZ], for sequences satisfying a technical condition (see also [E,KK2,KK1,HY]).

.., tn and s(n) = n.

Non-Følner Sequence Results

Corollary 4.7 states that if F is not a Følner sequence and 0 htop(T). p. 1

Where the paper says it

A generalization of Fekete’s Lemma, often called Ornstein-Weiss Lemma, states that when ϕ: F(G) → [0, ∞) is subadditive and translation-invariant then F 7→ 1 |F| ϕ(F) converges to a limit as F becomes more and more invariant. The next result describes what happens when we compute the limit along a sequence (Fn)n∈N which is not necessarily Følner, and we take the supremum over a family of functions ϕ with natural closure properties.

This result is derived from Theorem 4. p. 1
Where the paper says it

The next result describes what happens when we compute the limit along a sequence (Fn)n∈N which is not necessarily Følner, and we take the supremum over a family of functions ϕ with natural closure properties. 1 (Special case of Theorem 3.

6 using the constant C(F) defined in Definition 4. p. 3
Where the paper says it

One easily verifies that for E, F ⋐ G we have (U E) F = U EF . We also remark that the map F(G) → [0, ∞), F 7→ log N(U F ) is subadditive and (right) translation-invariant as defined in Definition 2.

2.

Goodwyn’s Theorem Proof Outline

The proof of Theorem 5.1 involves bounding Hµ(ξ F) using covering numbers and relating them to topological entropy via Proposition 5. p. 4

Where the paper says it

The measure-theoretic entropy hµ(T, ξ) of T with respect to ξ is defined as the limit of F(G) → [0, ∞), F 7→ 1 |F| Hµ(ξ F ) as F becomes more and more invariant. The limit exists by Proposition 2.

6. This leads to the inequality h Aµ(T, ξ) ≤ h Atop(T) when µ(∂ξ) = 0. p. 5
Where the paper says it

We prove that for all ϵ > 0 we have (2) lim sup n→∞ 1 s(n) ϕ(Fn) ≤ C(F, S) · (h + ϵ) + ϵ · ϕ({1G}). The claim in the statement follows from this inequality.

Sequence Entropy Bounds for Non-Følner Sequences

Lemma 3.2 provides a bound for the limsup of the sequence entropy involving C(F, S) and h, which is crucial for proving Theorem 3. p. 8

Where the paper says it

Furthermore, the same argument from before shows that the limsup of the bad part is at most lim sup n→∞ |{g ∈ EFh(n) : g bad}| s(h(n)) ϕ({1G}) With the same argument as before, our choice of E = RQ allows us to bound lim sup n→∞ |{g ∈ EFh(n) : g bad}| s(h(n)) ≤ lim sup n→∞ |RQFh(n) | s(h(n)) − lim inf n→∞ |QFh(n) | s(h(n)) Now, thanks to our choice of (h(n))n∈N, the first limsup is in fact a limit, so the last quantity is equal to = lim inf n→∞ |RQFh(n) | s(h(n)) − lim inf n→∞ |QFh(n) | s(h(n)) This value is at most ϵ by Equation (5), and this finishes the argument. □ Theorem 3.

3. The proof involves partitioning EFh(n) into good and bad parts. p. 6
Where the paper says it

This should become clear in the rest of the proof. We split the sum in the “good part” and the “bad part” as = lim sup n→∞   X i∈N (n,i) good 1 s(h(n))ϕ(EFh(n) ∩ Tigi) + X i∈N (n,i) bad 1 s(h(n))ϕ(EFh(n) ∩ Tigi)   We will now bound the good part from above.

Liminfs and Limsups Behavior

The paper discusses how liminfs and limsups behave differently when taking subsequences, requiring modifications to the argument for Equation (4). The final result for the infimum case is shown to be at most ϵ by Equation (5). p. 3

Where the paper says it

For fixed U the function F 7→ log N(U F )/|F| converges to a limit htop(T, U), but it is unknown if this limit equals the infimum of log N(U F )/|F| with F ranging over finite subsets of G. However, the following result is proved in [DFR, Theorem 6.

Classical Proportionality Results

Proposition 4.3 establishes that F is a (left) Følner sequence if and only if C(F) = 1. p. 2

Where the paper says it

Recall that a countable group is amenable if and only if for all ϵ > 0 and E ⋐ G there exists a finite subset of G which is (E, ϵ)-invariant. A sequence (Fn)n∈N of finite subsets of G is called a (left) Følner sequence when for every ϵ > 0 and E ⋐ G the set Fn is eventually (E, ϵ)-invariant.

This result connects the quantity C(F) to the notion of a Følner sequence. p. 0
Where the paper says it

AVERAGES OF SUB-ADDITIVE FUNCTIONS THROUGH NON-FØLNER SEQUENCES NICANOR CARRASCO-VARGAS Abstract. We study the quantity supϕ∈Φ lim supn→∞ ϕ(Fn)/|Fn|, where F = (Fn)n∈N is a sequence of finite subsets of a countable amenable group, and Φ is a collection of subadditive functions.

Topological Entropy via Sequence

Theorem 4.6 shows that sup U lim sup n→∞ log N(U Fn) Fn = C(F) · htop(T). p. 3

Where the paper says it

Proposition 2. The topological entropy htop(T) satisfies htop(T) = sup U inf F ⋐G 1 |F| log N(U F ) Next we define topological sequence entropy.

This result is derived by applying Theorem 3. p. 1
Where the paper says it

The next result describes what happens when we compute the limit along a sequence (Fn)n∈N which is not necessarily Følner, and we take the supremum over a family of functions ϕ with natural closure properties. 1 (Special case of Theorem 3.

3 with S = (s(n))n→N given by s(n) = Fn. p. 0
Where the paper says it

Introduction Sequence entropy is an invariant introduced by Kushnirenko [K] with the purpose of distinguishing pairs of zero-entropy dynamical systems. Given a sequence A = (ti)i∈N of natural numbers, one replaces the iterates T 1 , T2 , .

p. 0
Where the paper says it

Introduction Sequence entropy is an invariant introduced by Kushnirenko [K] with the purpose of distinguishing pairs of zero-entropy dynamical systems. Given a sequence A = (ti)i∈N of natural numbers, one replaces the iterates T 1 , T2 , .

Measure-Theoretic Entropy via Sequence

Theorem 4.10 shows that sup ξ lim sup n→∞ Hµ(ξ Fn) Fn = C(F) · hµ(T). p. 4

Where the paper says it

Let A = (ti)i∈N be a sequence of elements in G. The measure-theoretic sequence entropy h A µ (T, ξ) with respect to a finite measurable partition ξ is defined as h A µ (T, ξ) = lim sup n→∞ 1 n Hµ(ξ {t1,.

This is derived by applying Theorem 3. p. 1
Where the paper says it

The next result describes what happens when we compute the limit along a sequence (Fn)n∈N which is not necessarily Følner, and we take the supremum over a family of functions ϕ with natural closure properties. 1 (Special case of Theorem 3.

3 with S = (s(n))n→N given by s(n) = Fn.

Corollary on Non-Følner Sequences for Measure-Theoretic Entropy

Corollary 4.11 states that if F is not a Følner sequence and 0 hµ(T). p. 1

Where the paper says it

A generalization of Fekete’s Lemma, often called Ornstein-Weiss Lemma, states that when ϕ: F(G) → [0, ∞) is subadditive and translation-invariant then F 7→ 1 |F| ϕ(F) converges to a limit as F becomes more and more invariant. The next result describes what happens when we compute the limit along a sequence (Fn)n∈N which is not necessarily Følner, and we take the supremum over a family of functions ϕ with natural closure properties.

This result follows from Theorem 4. p. 1
Where the paper says it

The next result describes what happens when we compute the limit along a sequence (Fn)n∈N which is not necessarily Følner, and we take the supremum over a family of functions ϕ with natural closure properties. 1 (Special case of Theorem 3.

10 using the constant C(F) defined in Definition 4. p. 3
Where the paper says it

One easily verifies that for E, F ⋐ G we have (U E) F = U EF . We also remark that the map F(G) → [0, ∞), F 7→ log N(U F ) is subadditive and (right) translation-invariant as defined in Definition 2.

2.

Sequence Entropy Generalizations

The paper generalizes classical results of Goodman, Krug, and Newton to countable amenable groups by interpreting the proportionality principle as a statement about subadditive functions. This is achieved by removing technical conditions from previous generalizations. p. 0

Where the paper says it

The proportionality principle mentioned in the previous paragraph was first discovered by Krug and Newton [KN] for measure-theoretic sequence entropy, and subsequently generalized by Goodman [G] to topological sequence entropy. A generalization of the same results for actions of countable amenable groups was recently proved by Gao, Huang and Zhu [GHZ], for sequences satisfying a technical condition (see also [E,KK2,KK1,HY]).

Acknowledgements

Part of this work was done during the Simons Semester “Invariant Structures in Groups, Spaces, and Algebras”, at IMPAN, Poland, and the author acknowledges financial support from the Simons Foundation grant (award no. SFI-MPS-T-Institutes-000108408), and State Treasury funds as part of a task commissioned by the Minister of Science and Higher Education under the project “Organization of the Simons Semesters at the Banach Center - New Energies in 2026-2028” (agreement no. MNiSW/2025/DAP/491).

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 estimation of sequence entropy via non-Følner sequences: The system can now estimate topological or measure-theoretic entropy using a sequence Fn that is not necessarily Følner by obtaining an estimate of C(F) multiplied by h, as shown in Theorem 3.3: sup phi in Phi lim sup n→∞ phi(Fn)/Fn = C(F) · h.

  2. Generalization of proportionality principles: The AI system can now generalize classical results for sequence entropy to actions of countable amenable groups, removing prior technical conditions and interpreting the proportionality principle as a statement about subadditive functions.

  3. Enhanced handling of non-Følner sequences: For positive entropy cases, the system can explicitly quantify how much a non-Følner sequence overestimates topological or measure-theoretic entropy by the multiplicative factor C(F), where every choice of a non-Følner sequence overestimates topological and measure-theoretic entropy by the multiplicative factor C(F) (in the positive entropy case).

  4. Robust error bounding for variational principles: The system can rigorously bound topological sequence entropy relative to measure-theoretic entropy, proving that sup µ h Aµ(T) ≤ h Atop(T) with equality when conditions like K(A) 0 are met.

  5. Improved entropy estimation for non-Følner sequences: The system can compute the supremum of sequence entropies using a non-Følner sequence by obtaining the bound sup phi in Phi lim sup n→∞ phi(Fn)/s(n) = C(F) · h when C(F) 0.

Abstract

We study the quantity ϕ inΦ n to infinityϕ(F n)/F n, where F= (F n) n in N is a sequence of finite subsets of a countable amenable group, and Φ is a collection of subadditive functions. Under mild conditions on Φ we prove that the aforementioned quantity equals the supremum over ϕ inΦ of the limit of ϕ(F)/F as F becomes more and more invariant, multiplied by a ``Følnerness'' constant C(F) in[1, infinity]. We have C(F)=1 if and only if F is Følner. A corollary of this result is that every choice of a non-Følner sequence overestimates topological and measure-theoretic entropy by the multiplicative factor C(F) (in the positive entropy case). We obtain a generalization of classical results of Goodman, Krug, and Newton, on the so-called proportionality principle for sequence entropy, and their extension by Gao, Huang, and Zhu to countable amenable groups.

Sources

Related papers