Frequently dense harmonic functions and universal martingales on trees

Evgeny Abakumov; Massimo Picardello; Vassili Nestoridis

REVIEW 3 major objections 5 minor 15 references

Frequently dense harmonic functions and universal martingales on trees

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper proves that on trees with no infinite linear branches, for each prescribed sequence of levels, the harmonic functions whose restrictions at those levels are dense among measurable boundary functions form a dense Gδ set, and the…

desk verdict Real contribution to universality on trees, but the proof of Theorem 3.4 has a repairable gap: the stated extension lemma is too weak, though the paper's own construction supplies the needed stronger version. read the letter →

arxiv 1908.05579 v3 pith:VGYPC5LO submitted 2019-08-15 math.FA

Evgeny Abakumov , Vassili Nestoridis , Massimo Picardello This is my paper · ORCID

classification math.FA MSC 05C0531A2060J45

keywords harmonicfunctionsontreesboundarymartingalesuniversalfrequentuniversalitytransienttransitionoperatorsPoissontransformtreeapproximationinmeasure

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper establishes a discrete analogue of universal trigonometric approximation for harmonic functions on infinite trees. It shows that whenever the tree has no infinite linear branch, for any prescribed sequence of levels, the harmonic functions whose restrictions to those levels—viewed as locally constant functions on the boundary—approximate every measurable boundary function in measure form a dense Gδ set. If every vertex has enough descendants, the functions whose associated boundary martingale visits every open set of measurable functions with positive lower density are also dense. All these conclusions extend from the special forward-only transition operators to every very regular transient nearest-neighbor operator. A sympathetic reader would care because it shows that a single harmonic function on a tree can encode, in subsequences of its boundary values, the whole space of measurable functions, a universality phenomenon previously known for trigonometric and holomorphic series.

What carries the argument

Central machinery is the boundary-martingale representation of harmonic functions. A probability measure $\nu$ on the boundary $\Omega$ defines projections $\pi_n f$ that average $f$ over the arcs subtended by vertices of length $n$; these projections form a martingale, and the quotients $q(u,w)=\nu(I(w))/\nu(I(u))$ define a forward-only transition operator $Q$ whose harmonic functions are exactly the Poisson transforms of such martingales. The load-bearing estimate is the arc-shrinking inequality: from any vertex with no infinite linear branch, one can find a descendant path through branching vertices along which each chosen arc has measure at most half its parent's, so values at a sufficiently deep level approximate any target locally constant function within an arbitrarily small measure error. A harmonic-extension step then matches a prescribed function on an earlier ball by choosing a single exceptional value at the end of each path, and a category argument turns the dense approximants into a dense $G_\delta$ of universal functions. For frequent universality, an arithmetic lemma on indices $r_k$ whose levels with a given 2-adic valuation have strictly positive lower density supplies the schedule along which the martingale visits every open set with positive lower density.

What would settle it

A direct test of the dichotomy: take a tree that is a one-sided infinite chain, with a forward-only transition operator that moves only forward. Every harmonic function is constant along the chain, so the restrictions to circles are constant sequences and cannot be dense in the measurable functions on the boundary; hence U is empty, exactly as the theorem asserts. For frequent universality, construct a tree by inserting into every edge of a binary tree a linear chain of length $m_n$ growing without bound along the levels $n$; the boundary martingale then repeats values $m_n$ times in a row, so check whether the lower density of visits to any fixed open set drops to zero, which would confirm the role of the bounded-branch-length hypothesis in Theorem 3.4.

Watch

Extended reading notes

Core claim

On a rooted tree whose linear branches all have finite length, fix any forward-only transition operator Q, so that the random walk never moves toward the root. The paper proves that the set U(n) of Q-harmonic functions h whose level-n restrictions, lifted to locally constant functions on the boundary $\Omega$, form a dense sequence in the space of measurable functions on $\Omega$ is a dense $G_\delta$ subset of $H_Q$ for every sequence of levels n; and if the tree has an infinite linear branch, U is empty. When every vertex has at least two forward descendants, the set FU of harmonic functions whose boundary martingales visit every non-empty open set of measurable functions with positive lower density is dense in $H_Q$, while the set XU of functions whose visits have upper density one is a dense $G_\delta$ disjoint from FU, so FU is meager. The same statements hold for every very regular transient nearest-neighbor transition operator, that is, one with transition probabilities bounded below by $\delta>0$ and backward probabilities bounded above by $1/2-\delta$.

Load-bearing premise

The construction assumes that from every vertex one can descend through enough genuinely branching vertices to make boundary arcs shrink by a factor of at least two; this fails exactly on trees with an infinite linear branch, and frequent universality needs the stronger uniformity that every linear branch has bounded length.

Editorial extensions

If this is right

On any tree whose linear branches are all finite, universality is typical: for every choice of levels, the universal harmonic functions form a dense Gδ set, so almost every harmonic function in the category sense has a dense boundary-martingale orbit.
The same dense-Gδ conclusion holds for every very regular transient nearest-neighbor operator, so the phenomenon is independent of the precise transition probabilities provided they satisfy uniform lower and upper bounds.
Frequent universality, meaning visits with positive lower density, is dense but topologically negligible: the stronger upper-density-one set XU is residual and disjoint from FU, so the two notions of frequent behavior are genuinely different.
There is a dense vector subspace of HQ consisting, apart from zero, entirely of universal functions, giving algebraic genericity in addition to topological genericity.
If a tree contains one infinite linear branch, no universal harmonic function exists for forward-only operators, since harmonic functions are constant along such a branch.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The same arc-shrinking mechanism is quantitative: replacing the factor $1/2$ by any $\lambda<1$ should preserve all theorems, so the results should extend to transient operators whose Green kernel decays uniformly along branching paths, a larger class than the very regular operators named in Section 4.
A clean test suggested by Remark 3.5: on a tree obtained by inserting linear chains of length $m_n$ growing unboundedly at level $n$, universality should still hold while frequent universality should fail, because the boundary martingale repeats values for increasingly long stretches and the lower density of visits to any open set should drop to zero.
Read through the martingale representation, the same projection construction could be transplanted to any totally disconnected compact space with a tree-like filtration, for instance the $p$-adic integers, giving universal harmonic functions there as well.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Punchline: the paper is a genuine contribution to universality on trees, but the main frequent-universality theorem (Theorem 3.4) has a proof gap as written. The gap is repairable, so this is a revise-and-resubmit, not a reject.

What's actually new: full proofs of the dense G_delta and algebraic genericity results announced in [1], plus the frequently universal formulation with positive lower density, the meager FU vs XU dichotomy, and a transfer to very regular transition operators. The core construction in Theorem 3.1 is explicit and correct: solve the Dirichlet problem on balls, pick descendant paths with enough branching so the exceptional arcs have small measure, and use Baire category. The arithmetic lemma (Lemma 3.3) that builds the subsequence r_k is neat.

Soft spots:

1. Theorem 3.4 proof. The stated extension lemma only gives h*_{n+m} close to h*_n. Applied inductively, that makes h*_{r_k} Cauchy in measure, so the limit cannot be frequently universal. To get h*_{r_k} close to h*_{ell(k)} for a fixed dense sequence, you need the stronger lemma: any harmonic h on B_n extends to B_{n+m} with h*_{n+m} close to any prescribed K_{n+m}-measurable function. The construction in Theorem 3.1 actually provides this, because you can set the prescribed values on all but one subarc per direction and adjust the exceptional value to preserve harmonicity. Since ell(k) ≤ r_{k-1} eventually, h*_{ell(k)} is measurable at the required level, so the intended induction goes through. But the paper never states this stronger lemma, and the proof as written does not logically establish the theorem.

2. Corollary 4.1 (transfer to very regular operators) is too compressed. “It follows easily” is not enough for a claim that all of Section 3 transfers; a referee should ask for the Dirichlet extension argument spelled out with backward transitions.

3. Remark 3.5 (bounded linear branches) gives only a sketch, not a proof. If the claim is kept, it needs details.

4. Minor typo: in the Theorem 3.6 proof, “open sets in C” should be “open sets in M.”

Who it's for: people working on universality, harmonic analysis on trees, and boundary martingales. The self-citation to [1] is appropriate; this is the promised full version. The paper deserves a serious referee. I'd send it out and require a rewritten Theorem 3.4 with the stronger extension lemma, plus expansion of Section 4.

Referee Report

3 major / 5 minor

Summary. The manuscript studies harmonic functions on infinite locally finite trees with respect to nearest-neighbor transition operators. It identifies P-harmonic functions with boundary martingales via the Poisson transform, and proves several universality results. Theorem 3.1 states that, under the condition that every linear branch is finite, for any infinite sequence of radii n, the set U(n) of harmonic functions whose circle restrictions along n are dense in the space of measurable functions on the boundary is a dense G_delta subset of H_Q, and Theorem 3.2 adds algebraic genericity. Theorem 3.4, the main positive result, asserts that if every vertex has at least two descendants, then the set FU of frequently universal harmonic functions is dense. Theorem 3.6 claims that the set XU of functions whose visits to every open set have upper density 1 is a dense G_delta disjoint from FU, and Section 4 extends the results to very regular transient transition operators.

Significance. If correct, these results would provide a valuable discrete analogue of Menshov's universal trigonometric series and of frequent universality for holomorphic functions, with constructive proofs based on tree geometry and Baire category arguments. The frequent-density statement (Theorem 3.4) is the central new claim, and the paper also offers algebraic genericity and a meagerness complement. The approach via forward-only operators and boundary martingales is elegant. However, the present version contains two serious proof gaps: the induction in Theorem 3.4 uses an extension lemma too weak to force approximation of a prescribed universal target, and the Baire category representation of XU in Theorem 3.6 does not match the definition of XU. These gaps affect the main existence theorem and the meagerness theorem, respectively, and consequently also the transfer to very regular operators in Section 4.

major comments (3)

[Section 3, proof of Theorem 3.4] The proof claims that the argument in the last part of Theorem 3.1 yields an extension lemma of the form: any h on B_n harmonic in B_{n-1} extends to B_{n+m} with h^*_{n+m} in B(h^*_n, 2^{-m}). Even if this lemma is granted, the inductive application with n = r_{k-1} and m = ell(k) yields f^*_{r_k} in B(f^*_{r_{k-1}}, 2^{-ell(k)}), not the asserted f^*_{r_k} in B(h^*_{ell(k)}, 2^{-ell(k)}). The sequence f^*_{r_k} is therefore merely Cauchy in measure and need not follow the prescribed universal function h; the conclusion of frequent universality does not follow. The construction in Theorem 3.1 actually appears to support a stronger lemma in which an arbitrary A_{n+m}-measurable target replaces h^*_n, and the proof should state and prove this stronger lemma explicitly. It should also verify the measurability condition ell(k) <= r_{k-1} before applying the lemma, and it should show how, for each open set O, a fixed level m_0 with h^*_{m_0} in O together with Lemma 3.3 yields positive lower density of the hitting set.
[Section 3, Theorem 3.6, Eq. (3.9)] Theorem 3.6 defines XU as the set of h such that N(h,O) = {n : h^*_n in O} has upper density 1 for every non-empty open O. However, the sets E(j,m,n) in the proof are defined using the condition that there exist q > (1 - 1/m)n indices k_1 <= ... <= k_q <= n with f^*_{r_{k_l}} in O_j. Thus the intersection in (3.9) characterizes functions whose subsequence along the sparse set {r_k} visits O_j with upper density 1, not functions whose full sequence h^*_n does. Since the set {r_k} has positive but not full density in N, the equality (3.9) is false. The subsequent density argument, which refers to 'the same argument of Theorem 3.1', also appears to construct approximations only at the radii r_k and not at all indices n. The stated result that FU is meager in H_Q is therefore not established by the given proof.
[Section 4, Corollary 4.1] Corollary 4.1 asserts that Theorems 3.1, 3.2, 3.4 and 3.6 hold for every very regular transition operator. Since the proofs of Theorems 3.4 and 3.6 have the gaps described above, Corollary 4.1 inherits those gaps. In addition, the transfer itself is justified in a few sentences ('It follows easily that the results from Section 3 transfer without changes'); for a result that is part of the paper's advertised scope, the proof should spell out how the Dirichlet-problem construction of Theorem 3.1 is performed for a general very regular P, using the uniform bound U(v^-,v) <= 1 - epsilon from [10] to replace condition (3.3).

minor comments (5)

[Section 2.3 and proof of Theorem 3.1] In Definition 2 and in the proof of Theorem 3.1, the space M of measurable functions with metric d_nu should be described as the space of equivalence classes modulo nu-a.e. equality; otherwise d_nu is only a pseudometric.
[Section 3, proof of Theorem 3.1] The phrase 'Choose k > log2 s' should read k > log_2 s, and the condition should be 2^{-k} < 1/s, which is a minor typographical issue.
[Section 3, Theorem 3.6] The proof uses the symbol N both for the variable in quantifiers and for the set N(h,O), which makes the argument harder to follow; a different notation for the set of hitting indices would improve readability.
[Throughout] The manuscript contains several typographical errors (e.g., 'measu rable' in the abstract, 'of of' in Theorem 3.6) and would benefit from a careful proofreading pass.
[Section 4] The informal statement that the uniform decay condition is 'equivalent to an exponential decay of the hitting probability' would be clearer as a formal lemma quantifying the rate of decay obtained from the very regular bounds.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the universal-function constructions are direct existence arguments; self-citations in [1] and [10] are legitimate background, not reductions.

full rationale

The central results (Theorems 3.1, 3.2, 3.4, 3.6) are proved within the paper by explicit extension arguments. Theorem 3.1 fixes a test function f_j and constructs h by modifying finitely many values on a chosen level, solving the forward-only harmonicity equation for the exceptional value; it does not fit any parameter to the conclusion. Theorem 3.4 uses the dense G_delta set of universal functions from Theorem 3.1 and Lemma 3.3 to place f*_{r_k} near h*_{\ell(k)}; this is a transfer, not a definitional equivalence. The cited earlier work [1] supplies background on ramified sets but is not used to replace the proof. The bound from [10] used in Section 4 is parameter-free and concerns hitting probabilities of very regular operators, not the target universality; although it shares an author, it is independent evidence under the stated review rules. No equation is shown to equal its own input by construction, no fitted quantity is renamed as a prediction, and no uniqueness theorem is imported to force a choice. Accordingly there is no circular step. (A separate, non-circular issue: the induction in the proof of Theorem 3.4 as written appears to prove only convergence of f*_{r_k}, not frequent universality; this is a correctness concern, not a circularity one.)

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

No numerical constants are fitted; all hypotheses are qualitative bounds on branch lengths and transition probabilities. No new physical or mathematical entities are introduced beyond the existing tree, boundary, martingale, and harmonic function formalism.

assumptions (5)

domain assumption The tree T is locally finite, connected, simply connected, and has no terminal vertex.
Section 2.1: this is the standing definition of tree; all constructions use a fixed origin o and finite circles C_k.
domain assumption For forward-only Q, q(v-,v) is nonzero for every v different from the root, and for positive operators all transition coefficients are strictly positive.
Section 3 opening: zero-probability edges are pruned away, so the theorem is stated on the sub-tree where all forward jumps are possible.
domain assumption All linear branches of T have finite length for universality, and for frequent universality every vertex has at least two descendants, with a bounded linear branch variant in Remark 3.5.
These geometric hypotheses are necessary: an infinite linear branch makes U empty, and unbounded branch lengths can destroy frequent universality as explained in Remark 3.5.
domain assumption Very regular transition operators satisfy p(u,v) at least delta and p(u,u-) at most 1/2 - delta for some delta > 0, and are transient.
Definition 4 and Section 4: these uniform bounds, together with estimates from [10], replace the explicit arc decay condition (3.3) used for forward-only operators.
standard math Standard background results from [1], [5], [10], [11]: harmonic extension of prescribed values, Poisson integral representation, Fatou theorem, solution of the Dirichlet problem on finite balls, and the maximum principle.
Invoked in Sections 2, 3, and 4 as established theorems. Self-citation to [1] is heavy, but the new results do not reduce to [1] by construction.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Frequently dense harmonic functions and universal martingales on trees." pith.science (2026). https://pith.science/paper/VGYPC5LO

@misc{pith2026190805579,
  author       = {Pith},
  title        = {Pith review of: Frequently dense harmonic functions and universal martingales on trees},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VGYPC5LO}},
  note         = {Machine review of arXiv:1908.05579}
}

abstract

We prove the existence of harmonic functions $f$ on trees, with respect to suitable transient transition operators $P$, that satisfy an analogue of Menshov universal property in the following sense: $f$ is the Poisson transform of a martingale on the boundary of the tree (equipped with the harmonic measure $m$ induced by $P$) such that, for every measurable function $h$ on the boundary, it contains a subsequence that converges to $h$ in measure. Moreover, the martingale visits every open set of measurable functions with positive lower density.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 15 canonical work pages

[1]

Abakumov, V

E. Abakumov, V. Nestoridis, M. A. Picardello, Universal properties of harmonic functions on trees , J. Math. Anal. Appl. 441 (2017), 1181–1187

work page 2017
[10]

Kor´ anyi, M

A. Kor´ anyi, M. A. Picardello, M. H. Taibleson, Hardy spaces on non-homogeneous trees, Symp. Math. 29 (1988), 206–265

work page 1988
[2]

Atanasi, M

L. Atanasi, M. A. Picardello, The Lusin area function and local admissible conver- gence of harmonic functions on homogeneous trees , Trans. Amer. Math. Soc. 360 (2008), 3327–3343

work page 2008
[3]

Bayart, Universal radial limits of holomorphic functions , Glasgow Math

F. Bayart, Universal radial limits of holomorphic functions , Glasgow Math. J. 47 (2005), 261–267

work page 2005
[4]

Bendikov, A.A

A.D. Bendikov, A.A. Grigor’yan, Ch. Pittet, W. Woess, Isotropic Markov semigroups on ultra-metric spaces, Russ. Math. Surveys 69 (2014), 589–680 (Uspekhi Mat. Nauk 69, 3–102)

work page 2014
[5]

Cartier, Fonctions harmoniques sur un arbre , Symp

P. Cartier, Fonctions harmoniques sur un arbre , Symp. Math. 9 (1972), 203–270

work page 1972
[6]

Fatou, S´ eries trigonom´ etriques et s´ eries de TaylorActa Math

P. Fatou, S´ eries trigonom´ etriques et s´ eries de TaylorActa Math. 30 (1906), 335–400

work page 1906
[7]

Fig` a-Talamanca, M

A. Fig` a-Talamanca, M. A. Picardello, Harmonic Analysis on Free Groups , Lecture Notes in Pure and Applied Math. Marcel Dekker, New York, 1983

work page 1983

Show all 15 references

[8]

K. G. Grosse-Erdmann, Universal families and hypercyclic operators , Bull. Amer. Math. Soc. 36 (1999), 345–381

1999
[9]

J. G. K´ em` eny, J. L. Snell, A. W. Knapp, Denumerable Markov Chains , Graduate Texts in Mathematics 40, Springer-Verlag, Berlin–Heidelberg–New York, 1976

1976
[11]

M. A. Picardello, W. Woess, Finite truncations of random walks on trees , Appendix to the previous article
[12]

Kyrezi, V

I. Kyrezi, V. Nestoridis, Ch. Papachristodoulos, Some remarks on abstract universal series, J. Math. Anal. Appl. 387 (2012), 878–884

2012
[13]

Menshov, Sur les S´ eries Trigonom´ etriques Universelles, Comptes Rendus (Dok- lady) de l’Acad` emie des Sciences de l’URSS, XLIX (2), (1945), 79–82

D. Menshov, Sur les S´ eries Trigonom´ etriques Universelles, Comptes Rendus (Dok- lady) de l’Acad` emie des Sciences de l’URSS, XLIX (2), (1945), 79–82

1945
[14]

Papachristodoulos, Upper and lower frequently universal series , Glasgow Math

Ch. Papachristodoulos, Upper and lower frequently universal series , Glasgow Math. J. 55 (2013), 615–627

2013
[15]

Tor Vergata

W. Woess On the duality between jump processes on the ultrametric spa ces and ran- dom walks on trees , arXiv:1211.7216v1, Nov.30, 2012. LAMA, Universit ´ e Gustave Eiffel, Universit ´ e Paris Est Creteil, CNRS, F-77454, Marne-la-V all´ ee, France Email address : evgueni.abako...

2012 arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Abakumov, V

E. Abakumov, V. Nestoridis, M. A. Picardello, Universal properties of harmonic functions on trees , J. Math. Anal. Appl. 441 (2017), 1181–1187

work page 2017

[2] [10]

Kor´ anyi, M

A. Kor´ anyi, M. A. Picardello, M. H. Taibleson, Hardy spaces on non-homogeneous trees, Symp. Math. 29 (1988), 206–265

work page 1988

[3] [2]

Atanasi, M

L. Atanasi, M. A. Picardello, The Lusin area function and local admissible conver- gence of harmonic functions on homogeneous trees , Trans. Amer. Math. Soc. 360 (2008), 3327–3343

work page 2008

[4] [3]

Bayart, Universal radial limits of holomorphic functions , Glasgow Math

F. Bayart, Universal radial limits of holomorphic functions , Glasgow Math. J. 47 (2005), 261–267

work page 2005

[5] [4]

Bendikov, A.A

A.D. Bendikov, A.A. Grigor’yan, Ch. Pittet, W. Woess, Isotropic Markov semigroups on ultra-metric spaces, Russ. Math. Surveys 69 (2014), 589–680 (Uspekhi Mat. Nauk 69, 3–102)

work page 2014

[6] [5]

Cartier, Fonctions harmoniques sur un arbre , Symp

P. Cartier, Fonctions harmoniques sur un arbre , Symp. Math. 9 (1972), 203–270

work page 1972

[7] [6]

Fatou, S´ eries trigonom´ etriques et s´ eries de TaylorActa Math

P. Fatou, S´ eries trigonom´ etriques et s´ eries de TaylorActa Math. 30 (1906), 335–400

work page 1906

[8] [7]

Fig` a-Talamanca, M

A. Fig` a-Talamanca, M. A. Picardello, Harmonic Analysis on Free Groups , Lecture Notes in Pure and Applied Math. Marcel Dekker, New York, 1983

work page 1983