REVIEW 4 major objections 5 minor 24 references
Concentration of dimension in the Lagrange spectrum
T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash
Pith's one-line read For mixing horseshoes of conservative surface diffeomorphisms, the Hausdorff dimension of the Lagrange spectrum concentrates along a strictly increasing set J, with only countably many exceptional points.
desk verdict A well-structured proof of a genuine extension of [14], but the main theorem is conditional on three black-box results from the author's own preprints. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The carrying object is the cumulative dimension function $D_{\varphi,\Lambda,f}(t)=\operatorname{HD}(k^{-1}_{\varphi,\Lambda,f}(-\infty,t])$, tied to the horseshoe's unstable dimension by $D=\frac12 R_{\varphi,\Lambda,f}$. The proof's engine is a nested family of subhorseshoes $\tilde\Lambda_n$ chosen inside $\Lambda_{t_{n+1}-\delta_n}$ with $\operatorname{HD}(\tilde\Lambda_n)$ squeezed between $R(t_n)$ and $R(t_{n+1})$; these come from a subhorseshoe approximation result and an $O$-function/good-positions connection criterion, and are inserted into the level set $k^{-1}(\eta_-)$ by a Hölder homeomorphism $\Theta$ whose inverse is Hölder with exponent arbitrarily close to $1$. That transfer forces $\operatorname{HD}(K^u(\tilde\Lambda_0))\le\operatorname{HD}(k^{-1}(\eta_-))$ and yields the equality at every left endpoint. In the classical case the machinery is transplanted through the realization of initial segments $L\cap(-\infty,\sqrt{N^2+4N})$ as the dynamical spectra of horseshoes $\Lambda(N)$.
What would settle it
Take the horseshoe $\Lambda(4)$ that models $L\cap(-\infty,\sqrt{32}]$, and for a dense sample of $\eta\in(0,1)$ compute the left endpoint $\eta_-=\min\{t:D(t)=\eta\}$ together with $\operatorname{HD}(k^{-1}(\eta_-))$ via continued fractions with coefficients bounded by 4; the theorem predicts $D(\eta_-)=\operatorname{HD}(k^{-1}(\eta_-))$ at every sampled $\eta_-$, so a single discrepancy at a left endpoint would refute Theorem 4.1 and, by the transfer argument, Theorem 1.1.
Extended reading notes
Core claim
The central claim is Theorem 1.1: for every $C^2$ conservative diffeomorphism of a compact surface with a mixing horseshoe $\Lambda$, every $r\ge2$, and every $f$ in the open set of $C^r$ functions whose gradient is not perpendicular to the stable or unstable directions, $D_{\varphi,\Lambda,f}$ is continuous from $\mathbb{R}$ onto $[0,\operatorname{HD}(\Lambda)/2]$. Writing $\eta_-=\min\{t:D(t)=\eta\}$ and $J=\{\eta_-:\eta\in(0,\operatorname{HD}(\Lambda)/2]\}$, one has $D(t)=\operatorname{HD}(k^{-1}(t))$ for every $t\in J$, $D|_J$ is strictly increasing, and the dimension-carrying part of the spectrum is $J\cup F\cup\tilde J$ with $F$ countable and $D(t)\ne\operatorname{HD}(k^{-1}(t))$ on $\tilde J$. Section 4 transfers the theorem to the classical Lagrange spectrum, where $L\cap(3,\infty)=J\cup F\cup\tilde J$, $J=\{\eta_-:\eta\in(0,1)\}$, $D(J)=(0,1)$, and generically $L_{\mathrm{loc}}=L$ on $J$ while $L_{\mathrm{loc}}<L$ on $\tilde J$.
Load-bearing premise
The load-bearing premise is that two results from the author's own earlier preprints are correct: near each threshold there exist subhorseshoes whose Hausdorff dimension is arbitrarily close to the dimension of the half-line set, and any two such subhorseshoes connect inside the larger of their thresholds; if either premise fails, the nested family and the Hölder embedding on which Theorem 1.1 rests cannot be constructed.
Editorial extensions
If this is right
- For every $\eta\in(0,\operatorname{HD}(\Lambda)/2]$, the left endpoint $\eta_-$ satisfies $D(\eta_-)=\operatorname{HD}(k^{-1}(\eta_-))=\eta$, so the dimension-carrying levels are exactly these left endpoints.
- On the spine $J$, the function $D$ is strictly increasing and covers the interval $(0,\operatorname{HD}(\Lambda)/2]$, so the set where cumulative and level-set dimensions agree contains an uncountable interval-like family, with at most countably many extra coincidence points.
- For the classical Lagrange spectrum, $L\cap(3,\infty)$ splits as $J\cup F\cup\tilde J$, and at every point of $\tilde J$ the cumulative dimension $D(t)$ is strictly larger than the dimension of the level set $k^{-1}(t)$.
- Generically, when $\operatorname{HD}(\Lambda)<1$, local and cumulative dimensions of the dynamical spectrum agree on $J$ and disagree on $\tilde J$, making $J$ the part of the spectrum where local dimension is as large as possible.
- If the local-dimension function is non-decreasing on the derived set of $L$, then the exceptional set $\tilde J$ contains no accumulation points of the spectrum (Corollary 4.3).
Reading between the lines
- The paper notes that Theorem 4.1 does not settle injectivity of $D$ on the interior of the classical spectrum; a natural next question is whether the spine $J$ actually runs through every plateau of $D$, or whether some plateaus carry their dimension elsewhere in $\tilde J$.
- Because initial segments $L\cap(-\infty,\sqrt{N^2+4N})$ are modeled by the horseshoes $\Lambda(N)$, the identity $D(\eta_-)=\operatorname{HD}(k^{-1}(\eta_-))$ can be checked numerically for small $N$ using continued fractions with bounded coefficients, giving a direct test of the transfer from dynamics to number theory.
- The same connection-and-embedding mechanism is likely to work for any locally maximal hyperbolic set with local product structure and a real function transverse to the stable and unstable directions, not only for mixing horseshoes; if so, the three-part decomposition would be a general phenomenon for dimension functions over hyperbolic spectra.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies the dynamical Lagrange spectrum of a smooth conservative diffeomorphism with a mixing horseshoe. It defines, for each point in the unstable Cantor set, a quantity k_{φ,Λ,f} generalizing the best constant of Diophantine approximation, and analyzes the dimension function D_{φ,Λ,f}(t)=HD(k_{φ,Λ,f}^{-1}(-∞,t]). The main theorem claims that for every r≥2 and every f in R^r_{φ,Λ}, D is continuous, its image is [0,HD(Λ)/2], and the positive-dimension part of the Lagrange spectrum decomposes as J∪F∪J̃, where D|J is strictly increasing, D(t)=HD(k^{-1}(t)) on J, F is countable, and equality fails on J̃. Under an additional generic hypothesis with HD(Λ)<1, the paper also claims agreement of local and global dimension functions on J and strict disagreement on J̃. The corresponding statements are then transferred to the classical Lagrange spectrum. The proofs rely substantially on results imported from the author's preprints, most notably Proposition 1 of [13] and Proposition 3.6 of [15], which are neither stated in full nor proved here.
Significance. If the imported results hold in the pointwise form used, the theorem is a substantial contribution: it gives a complete description of the set of spectral points where the level-set dimension equals the half-line dimension and shows that this set is a strictly increasing arc. The Hölder homeomorphism construction in §3.3 is self-contained and carefully written, and the author is explicit about the generic assumptions in Theorem 1.3 and Remark 1.2. The main weakness is that the central theorem is conditional on unproved statements from three preprints by the same author and collaborators. The paper would be much stronger if the author either stated and proved those imported propositions or clearly restricted the main theorem to the hypotheses under which they are known to hold. As it stands, the result is plausible but not independently verifiable from the manuscript alone.
major comments (4)
- [§3.2] The proof of Proposition 3.7 begins by invoking 'proposition 1 of [13]' to produce, for arbitrary f in R^r_{φ,Λ}, subhorseshoes Λ_n ⊂ Λ_{t_{n+1}-δ_n} with HD(Λ_n) > R_{φ,Λ,f}(t_n) and max f|Λ_n > t_n. The precise hypotheses of that proposition are not stated, and [13] is a preprint by the same author. In this paper, Theorem 2.5 and Remark 1.2 are carefully phrased for residual sets, whereas Theorem 1.1 is claimed for every f in R^r_{φ,Λ}. The pointwise validity of Proposition 1 of [13] is therefore a live assumption. If it holds only under generic or residual hypotheses, or requires extra transversality conditions, then the existence of the sequence {Λ_n} and hence Theorem 1.1 is not established.
- [§3.2] Immediately after Proposition 3.6, the text says that since the function O takes only finitely many values, 'without loss of generality, we can suppose that Λ_n connects with Λ_m before max{t_{n+1},t_{m+1}} for any n,m'. A finite range alone does not imply that all pairs connect; this is an unstated subsequence reduction. The reduction needs to be written explicitly and checked against the requirements t_n → η_-, t_0 = η_- - ǫ, and R(t_n) < HD(Λ̃_n) ≤ R(t_{n+1}). The same issue affects the construction of the nested sequence {Λ̃_n} in §3.2 and hence the Hölder homeomorphism Θ in §3.3. As written, this is a gap in the proof.
- [§3.4] Proposition 3.10 asserts that if Λ̃_1 ⊈ Λ̃_2 are subhorseshoes, then HD(Λ̃_1) < HD(Λ̃_2), and it is attributed to 'corollary 3.9 of [14]'. This strict monotonicity is used to prove that D|J is strictly increasing, a central bullet of Theorem 1.1. The statement is not proved in the manuscript, and it is not a general property of arbitrary hyperbolic sets; it depends on the specific conservative horseshoe setting and on the dimension formula HD(Λ) = 2 HD(K^u(Λ)). If the cited corollary fails or has additional hypotheses, the strict-increase conclusion is unsupported. The author should either include a proof or provide the precise statement of the quoted corollary.
- [§4] The proof of Theorem 4.2 for t > t_1 uses the assertion from [14] that for any j in J̃(t,ǫ) one has HD(Λ̃_j) < 0.99, together with the estimate HD(M\L) < 1 from [24]. These are imported results. They are less central to Theorem 1.1 than the issues above, but they are load-bearing for the local-dimension statement in the classical case. The author should state the precise form of the needed bound and its hypotheses, or prove it.
minor comments (5)
- [Table of contents] The section heading 'Preliminares' should be 'Preliminaries'.
- [§3.4] The phrase 'modulus some countable subset' should be 'modulo some countable subset'.
- [§3.2] In equation (3.2), the expression 'log|Cu(Λ_{η_-}, r_0)|/r_0 - c_1 < 1.001 Du' is ambiguous; parentheses should make clear that c_1 is subtracted before forming the quotient or after, as intended.
- [§3.3] The phrase 'By the spectral theorem' should likely be 'By the spectral decomposition theorem', since the reference is to the decomposition of hyperbolic sets into basic pieces.
- [§3.4] The definition of F_{φ,Λ,f} as the set of points of J^0_{φ,Λ,f} isolated on the left is called 'enumerable' without an explicit argument; the countability is plausible because such points are separated by intervals with no points of J^0, but a short justification would improve clarity.
Circularity Check
Theorem 1.1 is proved through a long original construction, but its key subhorseshoe existence, connectivity, and strict-monotonicity inputs are quoted verbatim from the author's own preprints [13], [15], and [14], so the central claim is load-bearing on unverified self-citations rather than on independent support.
-
self citation load bearing
[Section 3.2, sequence of subhorseshoes (proof of Theorem 1.1)]
"Now, proposition 1 of [13] applied to t_{n+1} and η_n = 1 − R_{φ,Λ,f}(t_n)/R_{φ,Λ,f}(t_{n+1}) let us find δ_n > 0 and some subhorseshoe Λ_n ⊂ Λ_{t_{n+1}−δ_n} with HD(Λ_n) = 2HD(K^u(Λ_n)) > ... = R_{φ,Λ,f}(t_n) and in particular max f|Λ_n > t_n."
This is the sole source of the subhorseshoes Λ_n whose dimensions interpolate R(t_n) and whose f-maxima exceed t_n. These Λ_n feed Proposition 3.7, which in turn feeds the Hölder embedding Θ in §3.3 and the equality D(η_-)=HD(k^{-1}(η_-)). The proposition is not stated or proved in this paper, and it comes from the author's own preprint [13] (Moreira–Villamil–Lima). Since Theorem 1.1 claims the conclusion for every f∈R^r_{φ,Λ}, delegating the existence of these subhorseshoes to a same-author preprint without stating its hypotheses makes the central claim depend on an unverified self-citation rather than on a proof in this paper.
-
self citation load bearing
[Section 3.2, paragraph introducing O and Proposition 3.6]
"Using these results, in [15] is constructed a function O: N∪{0} → ⋃_{j=2}^{k−1} B^j_0, with the property that if for some m,n∈N∪{0} one has O(m)=O(n) then it is possible to go from Λ_m to Λ_n ... without leaving Λ_{max{t_{n+1},t_{m+1}}} ... Then, proposition 2.9 let us conclude: Proposition 3.6. Let m,n∈N∪{0} such that O(m)=O(n). Then Λ_m connects with Λ_n before max{t_{n+1},t_{m+1}}."
This connectivity statement is what turns the individual subhorseshoes into the nested sequence {Λ̃_n} via Corollary 2.10, with R(t_n)<HD(Λ̃_n)≤R(t_{n+1}) and max f|Λ̃_n<t_{n+1}. Proposition 3.7, the core of the proof, is therefore entirely conditional on the O-function/good-positions construction of [15], a preprint by the present author. No proof of the O-function's existence or of Proposition 3.6 is included, and the paper does not justify that the hypotheses of [15] hold for arbitrary f∈R^r_{φ,Λ}; the universal statement of Theorem 1.1 inherits this unproved dependence.
1 more flagged steps
-
self citation load bearing
[Section 3.4, before Proposition 3.10]
"Now, the spectral decomposition theorem and the corollary 3.9 of [14] let us conclude the following proposition: Proposition 3.10. Given two subhorseshoes Λ̃_1 and Λ̃_2 of Λ such that Λ̃_1 ⊈ Λ̃_2, we have HD(Λ̃_1) < HD(Λ̃_2)."
The strict increase of D on J, a required bullet in Theorem 1.1, is obtained from this proposition: the paper immediately uses 'As Λ̃_{n_k} ⊈ Λ̃_{n_k+1} for k∈N, this proposition let us conclude that ... D_{φ,Λ,f}|J^*_{φ,Λ,f} is strictly increasing.' The strict monotonicity of Hausdorff dimension for nested subhorseshoes is not proved here; it is delegated to corollary 3.9 of [14] (Moreira–Villamil), another same-author preprint. Thus a second clause of the main theorem rests on an imported self-citation that is not independently established in the present work.
full rationale
The paper contains a substantial original argument: the construction of Θ, the Hölder-inverse estimate, the decomposition into J, F, and J̃, and the transfer to the classical Lagrange spectrum are written in detail. However, the three load-bearing inputs—Proposition 1 of [13] (existence of subhorseshoes with prescribed dimension and max f), the O-function/good-positions construction and Proposition 3.6 of [15] (connectivity of those subhorseshoes), and Corollary 3.9 of [14] (strict dimension monotonicity)—are all quoted from preprints by the present author and are not proved or even fully stated here. These are not machine-checked, code-reproduced, or externally falsified in the paper, so under the review rule they do not count as independent support. The central claim does not reduce to its inputs by definition: the Θ embedding is a genuinely new step. But the theorem's universality ('for any r≥2 and f∈R^r_{φ,Λ}') is conditional on the unstated hypotheses of those same-author preprints, and a failure of any one of the imported results would invalidate Theorem 1.1. This is self-citation that is load-bearing rather than merely ancillary, so the circularity score is 4 rather than 0–2. There is no evidence of a fitted parameter renamed as a prediction, and the main statements are not definitionally equivalent to their inputs.
Assumptions & free parameters
assumptions (10)
- standard math Standard bounded distortion and exponential estimates for unstable intervals (Eqs. 2.1-2.3).
- standard math Existence of a Markov partition with small diameter and C^{1+α} stable/unstable foliations in a neighborhood of Λ.
- domain assumption Spectral decomposition of hyperbolic sets of finite type into subhorseshoes and transient sets, with dimension formulas (2.4) and (2.5).
- domain assumption Theorem 2.4: for f∈P^r, R_{φ,Λ,f}(t)=2Du(Λ_t) and t↦R is continuous.
- domain assumption Theorem 2.5: generically (residual Ũ) R=L=M for f∈R^r.
- domain assumption Proposition 1 of [13]: existence of subhorseshoes Λ_n inside Λ_{t_{n+1}-δ_n} with HD close to R(t_{n+1}) and max f > t_n.
- domain assumption O-function construction and 'good positions' from [15], yielding Proposition 3.6 that Λ_m and Λ_n connect before max{t_{n+1}, t_{m+1}}.
- domain assumption Corollary 3.9 of [14]: strictly nested subhorseshoes have strictly larger Hausdorff dimension (Proposition 3.10 here).
- domain assumption Moreira's Theorem 2.1 on images of products of non-essentially affine Cantor sets.
- domain assumption HD(M\L)<1, from [24].
Cite this review
Pith. "Pith review of Concentration of dimension in the Lagrange spectrum." pith.science (2026). https://pith.science/paper/MPZEFOZV
@misc{pith2026241116939,
author = {Pith},
title = {Pith review of: Concentration of dimension in the Lagrange spectrum},
year = {2026},
howpublished = {\url{https://pith.science/paper/MPZEFOZV}},
note = {Machine review of arXiv:2411.16939}
}
abstract
Let $\varphi$ be a smooth conservative diffeomorphism of a compact surface $S$ and let $\Lambda$ be a mixing horseshoe of $\varphi$. Given a smooth real function $f$ defined on $S$, we define for points $\eta$ in the unstable Cantor set of the pair $(\varphi,\Lambda)$, a generalization, $k_{\varphi,\Lambda,f}(\eta)$, of the best constant of Diophantine approximation for irrational numbers. We study the set of points $\eta$ for which the sets $k_{\varphi,\Lambda,f}^{-1}((-\infty,\eta])$ and $k_{\varphi,\Lambda,f}^{-1}(\eta)$ have the same Hausdorff dimension and when the Hausdorff dimension of $\Lambda$ is less than one, we describe generically the local Hausdorff dimension of the dynamical Lagrange spectrum, $\mathcal{L}_{\varphi,\Lambda,f}$, restricted to this set of points. Finally, we recover the same results for the classical Lagrange spectra.
Reference graph
Works this paper leans on
-
[14]
C. G. Moreira and C. Villamil. Concentration of dimension in extrema l points of left-half lines in the Lagrange spectrum, Preprint (2023) available at arXiv:2309.1 4646
work page 2023
-
[13]
C. G. Moreira, C. Villamil and D. Lima. Continuity of fractal dimens ions in conservative generic Markov and Lagrange dynamical spectra, Preprint (2023) availab le at arXiv:2305.07819
arXiv 2023
- [15]
-
[24]
Matheus, C., Moreira, C.G.: Fractal geometry of the complemen t of Lagrange spectrum in Markov spectrum. Comment. Math. Helv. , 95, no. 3, 593–633 (2020) Christian Camilo Silva Villamil: SUSTech International Ce nter for Mathematics, P. R. China. Email address : ccsilvav@sustech.edu.cn
work page 2020
-
[1]
Arnoux, Le codage du flot g´ eod´ esique sur la surface modulaire, Enseign
P. Arnoux, Le codage du flot g´ eod´ esique sur la surface modulaire, Enseign. Math. (2) 40, no. 1-2, 1994, 29-48
work page 1994
-
[2]
D. Lima and C. G. Moreira, Dynamical characterization of initials segments of the Mar kov and Lagrange spectra
-
[3]
Y. Pesin, Dimension Theory in Dynamical Systems: Contemporary Views and Applications, University of Chicago Press, 1998
work page 1998
-
[4]
C. Matheus, C. G. Moreira, M. Pollicott and P. Vytnova, Hausdorff dimension of Gauss–Cantor sets and two applications to classical Lagr ange and Markov spectra . https://arxiv.org/abs/2106.06572
Show all 24 references
-
[5]
Cerqueira, C
A. Cerqueira, C. Matheus, C. G. Moreira. Continuity of Hausdor ff dimension across generic dynamical Lagrange and Markov spectra. Journal of Modern Dynamics , 12:151–174, 2018
2018
-
[6]
D. Lima, C. Matheus, C. G. Moreira and S. Roma˜ na Classical and Dynamical Markov and Lagrange spectra, World Scientific , 2020
2020
-
[7]
Freiman, Diophantine approximation and geometry of numbers (The Mar koff spectrum) , Kalininskii Gosudarstvennyi Universitet, Moscow, 1975
G.A. Freiman, Diophantine approximation and geometry of numbers (The Mar koff spectrum) , Kalininskii Gosudarstvennyi Universitet, Moscow, 1975
1975
-
[8]
Lima and C
D. Lima and C. G. Moreira. Phase transtitions on the Markov and L agrange dynamical spectra. Annales de L’Institute Henri Poincar´ e (C), Analyse non-li neaire, 1–31,2020
2020
-
[9]
B. P. Kitchens. Symbolic Dynamics: One-sided, Two-sided and Countable Sta te Markov Shifts , Universitext, Springer, 1997
1997
-
[10]
McCluskey and A
H. McCluskey and A. Manning. Hausdorff dimension for horsesho es. Ergodic Theory and Dy- namical Systems , 3:251–260, 1983
1983
-
[11]
M. Shub. Global Stability of Dinamical Systems . Springer-Verlag, 1986
1986
-
[12]
A. Markoff. Sur les formes quadratiques binaires ind´ efinies. Math.Ann., 15:381–406, 1879
-
[16]
C. G. Moreira. Geometric properties of Markov and Lagrange s pectra. Annals of Math. , 188: 145–170, 2018
2018
-
[17]
Perron, ¨Uber die Approximation irrationaler Zahlen durch rational e II , S.-B
O. Perron, ¨Uber die Approximation irrationaler Zahlen durch rational e II , S.-B. Heidelberg Akad. Wiss., Abh. 8, 1921, 12 pp
1921
-
[18]
C. G. Moreira and S. Roma˜ na. On the Lagrange and Markov dyn amical spectra. Ergodic Theory and Dynamical Systems , Volume 37, Issue 5, August 2017, pp. 1570 - 1591
2017
-
[19]
Cusick and M
T. Cusick and M. Flahive, The Markoff and Lagrange spectra, Mathematical Surveys and Mono- graphs, 30. American Mathematical Society, Providence, RI, 1989. x+97 pp
1989
-
[20]
C. G. Moreira and J.-C. Yoccoz. Tangences homoclines stables p our des ensembles hyperboliques de grande dimension fractale. Annales Scientifiques de l’ ´Ecole Normale Sup´ erieure, 43:1–68, 2010
2010
-
[21]
C. G. Moreira. Geometric properties of images of cartesian pro ducts of regular Cantor sets by differentiable real maps, Mathematische Zeitschrift , (2023) 303:3. doi:10.1007/s00209-022- 03151-z
2023 doi
-
[22]
Palis and F
J. Palis and F. Takens. Hyperbolicity and Sensitive chaotic dynamics at homoclini c biifurcations: fractal dimensios and infinitely many attractors . Cambridge Univ. Press, 1993
1993
-
[23]
Ito, Number Theoretic expansions, Algorithms and Metrical observations
S. Ito, Number Theoretic expansions, Algorithms and Metrical observations. S´ eminaire de Th´ eorie des Nombres de Bordeaux, 1–27,1984. CONCENTRATION OF DIMENSION IN THE LAGRANGE SPECTRUM 27
1984
Reviewed August 12, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.